Support
Need help?
For questions about your fund's data, companies, metrics, or account access, contact the admin on your team. For technical questions, feature requests, or bug reports, reach out to Taylor Davidson at Hemrock or open an issue on GitHub.
Getting started
This platform is designed to automate the collection, parsing, and tracking of portfolio company reports. Instead of manually copying numbers out of emails and spreadsheets, the system uses AI to extract metrics from whatever format your companies send — emails, PDFs, Excel workbooks, slide decks, images — and writes the data directly into each company's metrics history.
The fastest way to get data flowing is to forward reporting emails to the inbound address shown in Settings. You can forward emails yourself, or give the inbound address to your founders or fund analysts and ask them to CC or send reports directly. Every email that arrives at that address is automatically parsed: the system identifies which company it's from, extracts the metrics you've defined, and flags anything it's unsure about for your review.
Not everything arrives by email. Files you download — PDFs, Excel workbooks, Word docs, PowerPoint decks, CSVs and images — can be added to a company’s documents through theImport page, and spreadsheets of metrics or investment history can be pasted there in bulk.
Once data starts flowing, the Portfolio dashboard shows every company, the Review queue catches anything that needs a human decision, and the Analyst is available on every page. The goal is to spend less time on data entry and more time on the analysis and conversations that matter.
The app is four products an admin turns on separately in Settings — Portfolio Reporting, Investment workflow (deals and diligence), LP Reporting and Fund operations (accounting, forecast, compliance) — so you only see what your fund uses. On a phone you can install it as an app from the browser’s Share or Install menu.
Start
Members land on Start: an Analyst chat across everything you can see, with quick actions for the common tasks — add a company, an investment or a vehicle, import documents, add a deal, issue a capital call, or declare a distribution. Ask a question or pick a task. Each entity’s Admin page carries the same actions for that entity, with a Capital calls card showing each open call’s received, outstanding and overdue, how many partners have paid, and who says they wired.
Self-hosting & license
Under the hood, the platform uses a database, authentication, file storage, inbound email processing, and an AI provider, with prebuilt integrations for several third-party services across the stack. Download it from GitHub and deploy it on your own accounts; DOCS.md in the repository is the full installation guide — database, environment variables, encryption, email providers, AI and deployment. You control the operational details and the costs. Most services in the stack have generous free tiers; AI features need an API key from Anthropic, OpenAI or OpenRouter, which are paid accounts.
License: Apache License 2.0. You are free to use it, modify it, and deploy it on your own infrastructure — for your own fund or commercially. There are no per-seat fees and no single-fund restriction. It includes an express patent grant, and the software is provided as-is, without warranty of any kind. Per its trademark clause, it does not grant rights to the “Hemrock” or “Unstructured Ventures” names or logos — if you fork or redeploy the software, please use your own branding. Read the full license.
Taylor Davidson of Hemrock offers paid setup and support on your own infrastructure and accounts, including onboarding your portfolio data — write to hello@hemrock.com.
Portfolio
The Portfolio page is the main dashboard for monitoring the fund. Companies show as cards or a table with their headline metrics (such as MRR and cash balance), filtered by status (Active by default) and sorted by name or investment date. Fund holdings and digital assets are listed below the companies.
An entity picker narrows the dashboard to one fund or SPV; your choice is saved to your account, and an admin sets the default under Settings → Portfolio Reporting → Dashboard entities. Settings → Default metrics applies a metric set to every company.
Click any company to open its detail page. Anyone with write access to the portfolio can edit company details, metrics and aliases (alternative names that help the system match inbound emails to the right company).
The header’s notes and chat buttons open the fund-level notes and the Analyst. Each note belongs to an entity, and is read by the people who can see that entity.
Company Detail
Clicking a company on the Portfolio dashboard opens its detail page. At the top you'll see the company name, headline metrics (like MRR and cash balance), and badges for stage, industry and portfolio groups — or Fund holding / Digital asset. Anyone with write access can edit the name, aliases, stage, industry, founders, overview, and other details that give the AI more context for analysis. The same page holds fund holdings (with the underlying fund’s register: commitment, NAV and capital calls) and digital assets (with wallets and units), and companies and assets show a price feed with any marks owed; it shows only the panels that kind of holding uses.
The main content area starts with the Analyst card. This is where you can generate an AI-powered summary of the company based on all available data — reported metrics, email content, uploaded documents, and any previous summaries. The AI acts as a senior analyst preparing a portfolio review memo: it highlights current performance, trends, strengths, risks, and follow-up questions. You can regenerate the summary at any time as new data comes in, clear it to start fresh, or upload additional context documents (board decks, strategy memos, investor updates) directly from this card to give the AI more to work with. If your fund has multiple AI providers configured, a provider selector lets you choose which AI to use for each generation.
Below the Analyst is the metrics section, where each metric has its own chart card. You can add new metrics directly from this page using the “Add metric” button, or delete a metric and all its data from its card. Each chart shows data points over time, color-coded by confidence level (green for high, amber for medium, red for low) with manual entries shown as hollow circles. Click any data point on a chart to open a popover where you can view the full details — period, value, confidence, source email, and notes — and edit or delete the value directly. You can also add data points manually using the “Add” button on each card, which is useful for entering historical data or correcting values. An export button at the top lets you download all metric data as a CSV. An Updates panel lists each update the company reported, with its period, attachments, an inline preview, and the OCR status of scanned pages.
Further down the page, a documents section lists all files associated with the company — both files you've uploaded and attachments from processed emails. These documents are available to the Analyst when generating summaries. Individual file uploads are limited to 20 MB per file. Finally, if the company has additional details like founders, contact emails, an overview, investment thesis, or a current business update, those appear at the bottom. A notes panel on the right side (or toggled via the chat button on mobile) lets your team leave company-specific observations. Each note belongs to an entity: on a company two funds share, Fund I’s notes are read only by people who can see Fund I.
The company page also has an Investments section with the fund’s transactions in that company — see Investments.
Review
When inbound emails are processed, the AI pipeline sometimes flags items that need a human decision. These flagged items appear in the Review queue. Common reasons include: a new company was detected that doesn't match any existing portfolio company, a metric value was extracted with low confidence, a reporting period was ambiguous or duplicated, the company couldn't be identified, or a metric couldn't be found in the report at all. Review also holds emails that match a deal in diligence (Diligence match, accepted into its data room by a person), and NAV, capital-call and distribution notices from funds you hold — each fund notice is approved into the right entity.
Each review item shows you the issue type, the extracted value (if any), and a snippet of context from the source email so you can make an informed decision. You can accept the extracted value as-is, reject it, or manually correct it with the right number. For new company detections, you can create the company or map it to an existing one.
The review badge in the sidebar shows how many items are waiting for attention. Once all review items for a given email are resolved, that email's status automatically moves from “needs review” to “success.” You can also dismiss all review items for an email at once if the entire report should be skipped.
Staying on top of the review queue is important — it's how you ensure the data flowing into your portfolio metrics is accurate. The system is designed to err on the side of flagging rather than silently writing bad data.
Approving changes
Changes the Analyst drafts rather than makes — metric updates, investments, capital calls — are approved or rejected in the Analyst panel, which also lists any still waiting from earlier conversations.
Inbound
Inbound shows every email that has been received and processed by the system. It's the audit trail for all automated report ingestion. Each row displays the sender, subject line, which company the email was matched to, and the processing status (success, needs review, failed, processing, pending, or skipped). Every email is first classified as reporting, an interaction, a deal, diligence, or other, and the detail page shows where it was routed.
Filter by status, company and date range (from the last 7 days to last year). Filters apply immediately as you change them. The list is paginated and sorted by most recent first, so new emails always appear at the top.
Clicking on any email opens its detail view. There you'll see the full processing result: which company was identified, which metrics were extracted and their values, the reporting period that was detected, and any review items that were created. The raw email body and attachment information are also available for reference.
If an email failed processing (for example, the AI key was misconfigured or the email content was unreadable), you can see the error message in the detail view. For emails stuck in “needs review,” you can open the review modal directly from the Inbound page to resolve flagged items without navigating to the Review queue.
The platform can also store documents for you automatically. If your admin has connectedGoogle Drive in Settings, every inbound email and its attachments are saved into company-specific folders — organized by company name — so you always have the original source files alongside the extracted data.
Email Detail
Clicking on any email in the Inbound list opens its detail page. At the top you'll see the subject line, sender address, received date, processing status badge, and the company the email was matched to (if identified). If processing failed, an error message explains what went wrong.
Below that, the page shows the metrics that were extracted — a table with each metric name, the reporting period, the extracted value, and a confidence indicator (high, medium, or low). If there are unresolved review items for this email, they appear next with their issue type, context snippet, and action buttons so you can accept, reject, edit, or dismiss each one. For new company detections, you can create the company directly from this page.
The detail page also lists any attachments that came with the email (with filename, type, and size), the raw email body text, and a collapsible view of the AI's full response for debugging or reference.
Approve accepts every open metric review on the email in one click (fund notices stay open for their own review). Process reruns the email — automatically (reclassified), or forced as reporting, an interaction or a new deal — replacing its extracted metrics and review items; you can also file it without processing, or skip it. If Google Drive is connected, Save to storage pushes the email and its attachments into the company’s folder, and a document can be uploaded against the email.
Company updates
Reporting emails become Company updates (Portfolio → Updates): a searchable record of what each company reported, with its attachments. Images and scanned PDF pages are queued for OCR — transcribed by the AI vision model — so their text is searchable too. You see the updates of the companies in your entities.
Import
Document upload matches each file to a company by name and saves it to that company’s documents, where the Analyst can use it. It accepts PDF, Word, PowerPoint, Excel, CSV and JPEG/PNG files up to 20 MB; files over 10 MB keep their extracted text only. To have a report’s metrics extracted, forward it to the inbound address instead.
You can also paste data that covers multiple companies at once — for example, rows copied from a spreadsheet or CSV file containing metrics across your portfolio. The system will parse the data, create new companies if they don't already exist, add new metrics as needed, and populate values for existing companies and metrics. An Email column adds those addresses to the authorized senders.
Additionally, you can paste investment transaction data — rounds, proceeds, valuations, and share prices — and the AI will parse the entries and match them to your portfolio companies. This is useful for bulk-importing cap table history, backfilling historical rounds, or onboarding an entire portfolio's investment data at once. Transactions are written to each company's Investments section automatically.
Asks
Asks lets you send reporting request emails to your portfolio companies. This is how you kick off a reporting cycle — compose a message asking companies to send in their latest numbers, select which companies should receive it, and send it out. The system tracks each request so you know what was sent and when.
The email composer takes a plain-text subject and body, pre-filled from your last ask. Set the from name and address, the reporting period and the date responses are due, add CC and BCC, and send yourself a test first. Emails go out through whichever outbound provider your admin has configured (Gmail, Resend, or Postmark).
Each request is logged with its recipient list, send timestamp, and delivery results. You can view past requests to see the full history of reporting asks. The response tracker is a grid of companies by quarter, each marked yes, no, n/a or waived, so you can see who still owes a report. Operational reminders (Settings) can email you when it’s time to send the quarterly ask and which companies haven’t responded by the due date.
When companies reply to your ask email with their report, those replies flow into the Inbound pipeline automatically (assuming the sender is on the authorized senders list and replies to the configured inbound address). The full loop — ask, receive, parse, review — is designed to work end to end with minimal manual effort.
Settings
Everyone can set their display name, two-factor authentication, note notifications, and their own API and MCP keys; non-admins can also connect their own Affinity key and read the AI summary prompt.
Admins configure the organization — fund name, logo and address, the sign-in page, appearance (accent colour), currency, team and roles, AI providers (Anthropic, OpenAI and/or OpenRouter), inbound email (Postmark or Resend) and authorized senders, outbound email (Gmail, Resend or Postmark), authentication email templates, file storage (Google Drive), analytics, usage tracking, operational reminders, Agent access and the version card — and turn on each product.
Each product group has a Turn on button and a row per feature, set to members, admins only, or off: Portfolio reporting (default metrics, the AI summary prompt, dashboard entities), Investment workflow (deal screening, external deal research, known referrers, AI models, diligence schemas, style anchors, defaults and caps, the Affinity connection), LP Reporting (the LP portal), and Fund operations — whose accounting configuration lives on each entity’s Admin page. A danger zone at the end of the organization settings permanently deletes all fund data.
Access
Two questions decide what a member sees, and both must say yes. Which areas: portfolio, deal flow, diligence, accounting, the management company, LP capital, GP economics, LP relations, compliance — each none, read, or read & write. Whose data: All entities (including ones added later, and items assigned to none), specific funds, SPVs and entities, or none. Set both in Settings → Team → Access; the first column is the member’s entities. Approving a join request asks which entities the person can see, and new members start with none. Admins always see everything.
On a company two funds hold, a Fund I member sees the company, its updates and KPIs, and Fund I’s position and books — not Fund II’s, nor that Fund II holds it. Deals and diligence records belong to an entity (unassigned ones are visible to admins until assigned); an LP shows only their positions in your entities; notes, documents and emails follow their company or entity; and the Analyst, MCP, the Claude and ChatGPT connections and API keys see exactly what the member sees. Something you cannot see is simply “not found”.
Feature visibility is the ceiling above all of it: a feature is on for members, admins only, or off — and off denies it to everyone, admins included, without deleting data.
For detailed technical setup instructions — configuring Supabase, environment variables, encryption keys, email providers, deployment, and more — see DOCS.md in the repository on GitHub.
Notes
Notes are on each company's detail page, on the Notes page, and in a notes panel on the Portfolio, Investments, Asks, Interactions, Compliance, LPs and deal pages. They provide a lightweight way for team members to share observations, context, and follow-up items without leaving the platform.
On a company detail page, the notes panel appears on the right side on desktop or can be toggled via a chat button on mobile. Notes here are specific to that company — use them for takeaways from founder calls, questions to raise at the next board meeting, context on a metric anomaly, or anything else your team should know about that particular investment. On the Portfolio dashboard, the shared notes section is for fund-level observations that apply across the portfolio: market trends, cross-company themes, reminders for the next investment committee, or general team updates.
The Notes page is a centralized feed that collects all notes across the fund in one place. You can filter by All notes, General (fund-level) notes, or just notes where you were @mentioned. Each note shows the author, timestamp, its entity, and which company it belongs to (if any), with unread notes highlighted so you can quickly catch up on what you've missed. Reply to a note to keep a thread together, and pin the ones that should stay at the top.
Notes support @mentions — type @ while writing a note to see a dropdown of team members, then select a name to mention them. Mentioned team members are highlighted in the note text and can receive email notifications depending on their preferences.
You can also follow companies to stay informed about notes posted on companies you care about, even if you aren't directly mentioned. When someone posts a note on a company you follow, you'll receive a notification.
Notification preferences are managed in Settings under your user profile. Choose All notes, @Mentions & followed companies (and pick the companies to follow), or None, which sends nothing.
All notes show the author's display name and timestamp. Team members can edit or delete their own notes. Each note belongs to an entity, and only people who can see that entity read it.
Interactions
Interactions gives GPs a searchable log of all conversations and introductions with portfolio companies. When a GP BCCs the fund's inbound email address on a conversation, the system automatically detects that the sender is a fund member, classifies the email as a CRM interaction (not a metrics report), and uses AI to extract a summary and identify any introductions.
The inbound classifier decides whether an email is a report, an interaction, a deal or something else; mail from fund members leans toward interactions. No manual tagging is required.
For each interaction, the AI generates a short summary, detects whether the email contains an introduction between parties, and extracts the names and context of anyone being introduced. Interactions are linked to portfolio companies when possible, so you can see all conversations related to a specific company.
The Interactions page shows all logged interactions across the fund. Filter by tag — intro, hiring, strategy, fundraising, product, partnership, legal, operations — for the tags present. Each entry shows the date, linked company, subject line, AI summary, and an intro badge when introductions were detected. Click the intro details to expand and see the names, emails, and context of introduced contacts.
On each company's detail page, a Recent interactions section shows the latest interactions for that company, with intro entries highlighted in a distinct style. A “View all” link takes you to the full interactions list.
The fund's inbound email address is displayed at the top of the Interactions page for easy reference and can be copied with one click. Simply BCC this address on any email conversation you want to log.
Deals
Deals is the inbound side of deal flow — cold pitches, partner-forwarded intros, and scout submissions arrive at your existing inbound email address and are screened against your fund's thesis before they reach a partner's inbox. Every inbound email runs through a content-aware classifier that decides between five destinations: reporting (portfolio metrics), interactions (CRM-style emails from fund members), deals (a company pitching the fund), diligence (an email about a company already in diligence), or other (newsletters, recruiter spam, vendor pitches — filed without processing). Sender identity is a strong signal but not a hard rule, so a partner forwarding a cold pitch lands in Deals where it belongs.
For each pitch routed to Deals, a single AI call extracts company name, founder, intro source, referrer when applicable, stage, industry, raise size, a 100–150 word company summary, and a thesis-fit analysis with a fit score (strong, moderate, weak, out of thesis). Out-of-thesis pitches auto-archive and surface in a weekly digest email so partners can sanity-check without eyeballing every cold pitch. Founders can also submit pitches directly via a public form at a per-fund URL — admins generate or rotate the URL in Settings. New deal adds one by hand, and each deal belongs to an entity. External deal research runs web research on the founders and company for deals at or above a fit threshold you set in Settings.
The Deals page lists active pitches as a sortable table or a kanban board (drag-and-drop across status columns New, Reviewing, Advancing, Met, Diligence, Invested and Passed); moving a deal to Diligence opens a diligence record pre-filled from it. Click a pitch to see the summary, thesis-fit analysis, source email, attachments, founders, intro source, and a deal-scoped Analyst chat that knows the pitch and your thesis.
Settings → Investment Workflow controls the investment thesis, screening prompt, public submission token, and the Known Referrers list (scouts and friends-of-fund whose intros bias toward Deals). Uncertain items go to a Review queue with the top two predicted destinations for one-click resolution — nothing is silently dropped.
Diligence
Diligence is the pre-investment workflow: when a deal is worth real time, you create a diligence record, upload the data room, and run a schema-driven agent that ingests the documents, conducts external research, asks partner Q&A, drafts a structured memo, scores it per your rubric, and renders to Word or Google Docs. Each diligence record has Checklist, Data Room, Research, Founders, Scoring, Memo and Settings tabs. The checklist is assessed against the data room (found, partial, missing, n/a); documents are categorized; call recordings are transcribed; notes and files can be imported from Affinity; inbound emails about the deal can be accepted into its data room; and each deal shows its token, cost and time usage.
The agent is operated by seven YAML/MD configuration files (“schemas”) that admins edit per-fund through an in-app editor under Settings → Investment Workflow → Diligence → Schemas: instructions, rubric, qa_library, data_room_ingestion, research_dossier, memo_output, and style_anchors. The schema editor is a plain-text editor with inline YAML syntax validation and version history; rolling back to a prior version is one click. Defaults are seeded automatically the first time you open the editor, so you can run the agent immediately and customize as you go.
Style anchors are uploaded reference memos that teach the agent your firm's voice. Upload 3–8 prior memos in Settings → Investment Workflow → Diligence → Style anchors, tag each with vintage, sector, voice representativeness, and partner notes, and the agent uses them to match structure and tone during drafting. Reference memos teach voice — they never supply facts to a new memo.
The agent runs in six stages: Ingest (classify each doc, extract claims, run gap analysis), Research (verify or contradict claims, build a competitive map, compile founder dossiers), Q&A (next-best partner questions per the qa_library with skip logic against ingestion + research), Draft (assemble paragraphs with paragraph-level citations), Score (rate each rubric dimension; partner-only dimensions like team get null score with supporting material), Render (markdown, .docx download, or native Google Doc). Long stages run as background jobs picked up by a worker every minute.
The memo editor is a two-pane view: rendered memo on the left with inline citation markers and visual treatment for projections, unverified claims, and contradictions; paragraph inspector and partner-attention sidebar on the right. Partners edit any paragraph (it flips topartner_edited origin), update rubric scores by hand, work through the attention queue (must-address / should-address / FYI), and finalize when ready — finalizing locks the draft. Recommendation and team scoring are partner-only and can never be set by the agent.
Across all your active deals, the Inbox aggregates open partner-attention items so you can triage them in one pass. The Analytics view shows the agent funnel (created → ingestion → research → Q&A → draft → finalized → won) with drop-off percentages, time-in-stage medians, win/loss by sector, and throughput per lead partner. Settings → Investment Workflow → Diligence → Defaults & caps sets per-deal and monthly token caps (with a current-month usage bar), the research web-search toggle and the transcription check; Per-stage AI models picks a model per stage — e.g. a cheap model for ingest, a stronger one for draft.
Investments
Investments lists every holding — companies, fund holdings and digital assets — at cost and value, with a per-vehicle summary (MOIC, IRR and, where the books exist, TVPI, DPI and RVPI) and two charts. One Add menu creates a company, fund holding, digital asset or vehicle. Filters narrow it by vehicle type, status and individual entity, and each vehicle’s name opens its entity page.
The holdings chart switches between Value (fair value against current cost) and Multiple (gross multiple — proceeds plus fair value over invested — against 1.0x). The multiple view follows the Status filter, so All includes exits and write-offs.
A direct deal or SPV whose holdings carry no figures yet — none recorded, or set up but never given an investment, proceeds or a mark — is shown as a Deal vehicle at the sum of its LP positions: paid-in as invested, distributions as proceeds, NAV as value. The moment any of its holdings has a figure, the holdings take over, so nothing is counted twice.
On each company detail page, the Investments section holds the fund’s transactions with that company: investments, conversions, proceeds and escrow receipts, valuation updates, rounds, share splits and income. Each one posts its entry to the entity’s books. Summary metrics — total invested, current FMV, MOIC, and total realized — are shown above the transaction table.
Anyone with write access to investments can add, edit or delete transactions. For exited companies the FMV reflects total realized proceeds; for written-off companies it shows zero; and for active companies it uses the latest share price multiplied by the shares still held — or, where no share price has been recorded, the remaining cost plus its marks. Investment data can also be bulk-imported via the Import page by pasting transaction data from a spreadsheet.
Entities & accounting
Entities lists every fund, SPV, GP entity, individual and management company, with performance — committed, called, distributed, NAV, DPI, TVPI, IRR — derived from the books, and a Net to LP lens. Open one for its pages: journal, bank, capital accounts, schedule of investments, statements, forecast, tax and close.
Accounting is always on
Every entity keeps double-entry books from the day it is created. Recording an investment, an exit, a mark, a conversion or a capital call posts its entry; a bank row matching one is linked to it rather than booked twice; and each income or expense entry is allocated to the partners as it posts. So the books are complete whether or not anyone opens them. The Accountingswitch (Settings → Feature visibility, or Turn on Fund Operations) only decides whether the pages are shown — like Investments and LP capital tracking.
Bringing in history: on the entity’s status page, Put them on the ledgerbooks investment history already in the tracker; Migrate from QuickBooks imports a general ledger; the bank page takes a CSV or a table copied straight from your bank’s website; andOpening balances takes over at a cutover date. An administrator’s statement can be kept beside the books as dated positions and reconciled against them.
Close reviews and locks each month: drafts and suggested recurring entries can be posted right there, it reconciles ledger cash to the bank feed, trues up carried interest, and lists every partner allocation in the month. The full guide is ACCOUNTING.md in the repository.
GP economics (carry terms, carry accrued and paid per partner, per-deal carry and GP entity ownership) and Tax reporting (Schedule K-1 packages per vehicle and tax year, the allocation behind them, and delivery to the LP portal) are separate switches, so someone can keep the books without seeing the partners’ carry.
Capital calls and distributions are issued from an entity’s Capital accounts (or its Admin page): as an amount or a percentage of commitments, split pro-rata, or pasted from a spreadsheet — a partner column, an amount, and optionally what was paid and when. The sheet fills each partner’s line for you to check, and anything already paid is recorded as a receipt when you issue, so the call shows who has paid from the start.
Forecast
One plan of the whole entity, built on the posted books and never written to them: its key figures on top (forecast gross MOIC, LP TVPI and IRR, distributions and carry to come) and four tabs — Charts (cash flows and multiples by year with their simulated ranges, revenue and expenses, cash, committed capital usage, simulated outcomes), Investments (current and planned deals, each with its hold period and forecast multiple), Capital planning (calls, fees and expenses, and investment, actual and forecast, over the selected dates) and Financials (income statement and statement of cash flows on an accrual or cash basis, the balance sheet at the end of each period, and journal entries). The date range and month/quarter/year apply on every tab. A management company has no portfolio, so it gets Charts, Financials and Capital planning, which there shows its cash: income received, expenses paid, money to and from its funds and owners, ending cash and the lowest point in the range.
Investments and exits are entered on Investments; fees and expenses are rules on Financials (any month can be overridden); calls and distributions follow a policy (a cash reserve, a payout lag) with any month set by hand, and the waterfall works out the carry on what is distributed. Scenario saves the plan as a scenario or as the budget actuals are compared against, copies it as a new scenario, or drafts one with the Analyst. Settings hold the return assumptions, the calls and distributions policy, and entity links — which fund pays which management company its fee.
The Analyst can change any of it — deals, exits, return assumptions, fees, calls, scenarios — as drafts you approve; a connected agent (MCP) can do the same directly with write access.
Management company
The management company is the firm’s own operating entity — the company that employs the team, collects the management fee and pays the rent — as opposed to the funds it manages. It is listed on Entities with the funds but in its own table, because it has no commitments, no NAV, no TVPI and no limited partners. Switch it on under Settings → Fund Operations, then add a vehicle of type Management company from Investments or Start: its chart of accounts is there from creation. Members also need the separate Management company grant (Settings → Team → Access).
Its chart of accounts is its own. Cash, receivables, prepaid and fixed assets, payables, payroll liabilities, deferred fee revenue, members’ capital and draws, fee and reimbursement income, and an operating expense structure that begins with four separate compensation accounts — salaries, payroll taxes, benefits, and bonus. They are kept apart because “what does a head cost us” and “what did we pay out on last year’s performance” are different questions, and once they are booked together neither can be answered from the ledger again.
The dashboard answers the four questions a firm asks about its own operating entity: how much cash there is and how long it lasts (burn excludes depreciation, which is not money leaving the building); what came in and went out by quarter, with the empty quarters drawn rather than skipped, because a fee that never arrived should look like a hole and not a flat line; where the money goes, ranked by account; and who owes whom.
Intercompany. Recording a charge — a management fee, an expense reimbursement, an allocated cost, an advance — posts both sides in one action: the receivable and the fee income on the management company, the expense and the payable on the fund, linked so the two can be reconciled. Settlement is a separate step, because the cash usually moves in a different quarter from the charge. Balances are shown per counterparty with due-from and due-to kept apart rather than netted, and they are read from the ledger rather than summed from the register — so a correcting journal entry is reflected without anyone having to amend a record as well. The same charges can be recorded from the fund side, on a fund’s Admin page.
The books themselves are the same tools a fund uses, scoped to this entity: the journal, the bank feed, financial statements, the period close, and the QuickBooks general-ledger import. The import’s account mapping knows the management-company vocabulary, so payroll, benefits, occupancy and technology accounts arrive already matched rather than needing a dropdown each.
Access is granted separately. A management company’s ledger carries salaries, bonuses and partner draws, and none of that appears anywhere in a fund’s books — the fund sees one number, the fee it pays. So Management company is its own access grant, it starts at no access for everyone, and granting someone fund accounting does not hand them the partners’ compensation. The rule is enforced by the entity rather than the page: an accounting request for a management company is refused whatever the caller’s accounting grant says.
Letters
Letters helps you generate quarterly update letters for your limited partners. Using AI and your portfolio data — reported metrics, company summaries, investment performance, and team notes — the system drafts professional LP communications scoped to a specific portfolio group and reporting period. To create a letter, click “New letter” and select the year, quarter, portfolio group, and template. You can optionally toggle “year-end summary” for Q4 letters and add custom instructions to guide the AI. A preview step shows the companies and data that will be included before generation begins.
If you've written LP letters before, you can upload a previous letter (.docx or .pdf) and the AI will analyze it to match your writing style, tone, and structure. Otherwise, a built-in default template is available. Templates are reusable across letters and managed from the Templates dialog on the Letters page.
During generation, the AI writes a narrative for each company in the portfolio group, drawing on reported metrics, recent trends, company summaries, investment data, and team notes. A portfolio summary table with investment performance is also generated, and the full letter is assembled from these sections. After generation, the letter opens in an editor with two views: “Sections” shows each company narrative individually for targeted editing, and “Full” shows the complete assembled letter. You can edit narratives inline, regenerate individual company sections or the entire letter, and add custom prompts per-company or globally to refine the output. Per-company prompts can either add to or replace the default generation prompt.
When you're satisfied with the letter, download it as .docx, or export it to Google Docs if Google is connected. With the LP portal on, Share with LPs publishes it to the portal.
LPs
LPs rolls every LP across every vehicle up to the investor, live as of any date: commitment, paid-in, distributions, NAV, DPI, TVPI and IRR, with charts, search and a vehicle filter. Print investor report cards (one PDF each, or all at once), export to Excel, group or rename investors, and set the report header and footer from its Settings button.
Capital accounts keeps each vehicle’s administrator statements as dated positions you paste or type, with history and inline editing. Sharing with LPs freezes a snapshot of the figures.
Documents manages the portal side: invite LPs and their authorized users (one at a time or from a pasted sheet), the onboarding checklist with LP uploads to review, shared documents, and messages. Preview portal shows the portal as an LP sees it, and Activity shows who logged in, viewed or downloaded.
These are separate switches under LP Reporting: LPs, LP Capital Accounts, LP Documents and Sharing, and LP Activity Log.
LP portal
Your LPs sign in to their own portal, separate from the app: an overview of their positions, notices for capital calls and distributions (with a “We’ve wired” form), letters, documents, onboarding, a way to contact you, and their settings. Each LP sees only their own entities.
Compliance
Compliance helps fund managers stay on top of regulatory filings, tax deadlines, internal compliance requirements, and fund reporting obligations. It provides a calendar-based view of everything due throughout the year, tailored to your fund's specific profile and registration status.
Start by completing a short compliance profile questionnaire about your fund — registration status, AUM range, fund structure, Reg D exemption type, state presence, and a few other details. The system uses your answers to automatically determine which compliance items apply, which need further review, and which you can safely dismiss. If your fund profile changes, update the questionnaire and the calendar adjusts accordingly.
All applicable items appear in a monthly calendar organized by deadline. Items are color-coded by category — SEC filings in amber, tax filings in green, internal compliance in blue, fund reporting in purple, state compliance in rose, CFTC in orange, and AML/FinCEN in red. Quarterly items like partnership expense reviews and access person disclosures appear in each quarter independently, so you can track and dismiss them separately. Items can be per vehicle, and you see those for your entities; other years are a click away, and upcoming filings appear in the operational-reminders digest.
When you finish a filing, mark it complete (with an optional note and filing link), or dismiss it if it doesn't apply; filter by Active, Completed, Dismissed or All. The page has Calendar, All Items, Fund Profile and Filing Links tabs. The all items view provides a comprehensive list of every compliance item organized by category, with each item showing its frequency, deadline, applicability, filing system, and any relevant notes or alerts.
The system ships with a curated registry of compliance items covering SEC filings (Form ADV, Form PF, Form 13F, Schedule 13G, Form 13H, Form N-PX), securities offerings (Form D, Blue Sky), CFTC exemptions, California diversity reporting, tax filings (Form 1065, K-1s, Form 7004), internal compliance (access person disclosures, annual compliance review, privacy notice), AML/FinCEN requirements, fund reporting (quarterly financials, valuations), and partnership expense allocation.
Save links to filing portals, regulatory accounts, and reference documents alongside your compliance items. Each link can optionally be associated with a specific compliance item, and associated links appear on that item's detail card for quick access during filing season. Links are editable after entry — hover over any link and click the pencil icon to update the title, URL, description, or associated compliance item.
Usage
AI Usage is an admin-only page that shows how your fund is consuming AI tokens and how team members are using the platform.
The top section shows AI token usage by provider (Anthropic, OpenAI and/or OpenRouter), with month-to-date input tokens, output tokens and estimated cost. A daily table breaks usage down by provider and model, and a monthly summary shows totals by month.
The bottom section shows team activity: logins, companies, imports and other actions per team member, and a recent activity feed. Activity logging can be turned off in Settings → Usage tracking.
Analyst
The Analyst is an interactive chat interface available on every page. Powered by AI, it acts as a senior venture capital analyst with full access to your portfolio data, answering questions, surfacing insights, and helping you prepare for board meetings and investment committee discussions.
On a company page, the Analyst has access to that company's reported metrics, email content, uploaded documents, previous AI summaries, investment transaction history, portfolio peer comparisons, and your team's internal discussion notes. You can ask it to analyze performance trends, compare the company to peers, identify risks, draft or refine summaries, interpret financial data from reports, or answer any question about the company's data.
Elsewhere it is scoped to the section you’re in — portfolio, deals, diligence, LPs, or an entity’s books — and only to the entities you can see; on Start it covers all of them. Use it to compare companies, get portfolio-level insights, or ask about cross-portfolio trends and themes.
Your chat history is persistent and saved to your account in the database. You can close the panel, navigate to other pages, or close the browser entirely — when you return, click the clock icon to open your conversation history and resume any previous thread. Conversations are scoped by context: company-specific chats stay with that company, and portfolio-wide chats have their own history.
The Analyst has conversation memory that gives it continuity across sessions. When you start a new conversation, the system automatically summarizes your recent past conversations in the same context and injects those summaries into the AI's prompt. This means the Analyst remembers what you've discussed before — key questions, conclusions, and concerns — without you needing to repeat context.
The Analyst also incorporates your team notes into its analysis. Notes posted on a company page are included when chatting about that company, and portfolio-wide notes are included in fund-level conversations. This means the AI is aware of your team's observations, follow-up items, and qualitative context alongside the quantitative data.
Use the header controls to manage conversations: the clock icon opens your conversation history, the plus icon starts a new conversation, and you can delete old conversations from the history list. Pick a model from any configured provider (or Auto) and, where the model supports it, an effort level.
The Analyst can also save its responses directly as company summaries using the “Save as Summary” button that appears below each response on company pages. This lets you use the chat to iteratively refine a summary and then commit it to the company's record with one click.
Attach a document (PDF, Word, Excel, Markdown, CSV or text) to ask about it; in an entity’s books the Analyst drafts the journal entry from a capital-call notice, invoice or wire, saved as a draft for you to review. Changes it proposes elsewhere — metric updates, investments, capital calls — appear as Approve/Reject cards in the panel, which also lists any still waiting from earlier conversations. On Forecast, Draft with AI opens it with the entity’s history.
Claude & ChatGPT
Use your fund from Claude or ChatGPT: ask questions, and open live portfolio, company, financial statement and LP dashboards inside the conversation — signed in as you, seeing only what you can see. An admin first switches on Settings → Agent access; then each member finds their connection under Settings → API and MCP.
Claude
- In Settings → API and MCP, download Plugin for Claude.
- In Claude, open Customize → Plugins → Add → Upload plugin and choose the file.
- Open the plugin’s Connectors tab, choose Connect, sign in to your fund and approve.
Claude Code: claude mcp add --transport http fund <MCP URL>and sign in when prompted, or add an API key header from the same Settings page.
ChatGPT
- Copy the MCP URL from Settings → API and MCP.
- In ChatGPT, open Plugins, choose +, then Add custom MCP server; paste the URL, upload the icon from Settings → API and MCP, and create it as a plugin.
- Connect, sign in to your fund and approve.
For the skills in the ChatGPT desktop app or Codex, download Plugin for ChatGPT, unzip it into ~/.agents/plugins/, list it in~/.agents/plugins/marketplace.json and restart the app (DOCS.md has the exact entry).
Getting started
Once connected, ask “What can you do?”. That opens your home dashboard: tiles for the dashboards your access allows (the portfolio, statements per entity, LP capital), the dashboards you and your colleagues saved, and questions for each area you can see — click one and it is asked for you.
Prompts are ready-made requests in your assistant’s prompt menu (in Claude, the + menu; in Claude Code, type /): Get started, Portfolio review, Company check-in, Quarter-end review, LP capital status, Draft a forecast and Budget variance — only those your access covers. Pick one, fill in the company or fund, and send. Ask for an LP’s capital account statement or an investor’s report card as a PDF and the assistant gives you a download link that works for an hour, for you only.
- Portfolio — “Show me the portfolio.” · “Which companies are marked below cost?”
- Books — “Pull up the balance sheet for our main fund, last quarter.” · “What were our largest expenses this year?”
- LP capital — “Who has funded their capital calls?” · “Who still owes us on the latest capital call?”
- Forecast — “Draft a 12-month forecast for the management company.” · “What are the largest variances to budget this quarter?”
- Deal flow — “What new deals came in this week?” · “Which inbound deals best fit our thesis?”
- Diligence — “What's still open on our diligence checklists?” · “Summarize the evidence on our current deal.”
The plugin is built for your own deployment’s address and contains no key and no data. You choose read, or read and change, when you approve; revoking it or switching Agent access off stops it at once. Ask the assistant to save this dashboard to reopen it later, in either assistant, always on current figures.
What the assistants can show
- Portfolio — pick an entity, switch between current holdings and all companies, and chart by fair value or by gross multiple.
- Capital calls — “who still owes us on the latest call?” shows each LP as paid, partly paid, says wired (their word from the portal, not yet received), unpaid or overdue, with the fund’s other calls to switch between.
- LP documents — ask for an LP’s capital account statement or an investor’s report card as a PDF; the link works for an hour, only for you, and checks your access again when opened.
- Carried interest — the fund’s carry is shown to anyone who can see LP capital; what each carry recipient personally earns needs GP economics access.
The connector’s icon. ChatGPT uses the icon you upload when adding the server (download it from Settings → API and MCP). Claude currently shows a generic icon for every custom connector; the server already declares your fund’s icon for when it reads it.
File Handling & Security
The platform accepts PDF, Word (.doc/.docx), PowerPoint (.ppt/.pptx), Excel (.xls/.xlsx), CSV and JPEG/PNG images. Files can be uploaded through the Import page, attached to inbound emails, or uploaded directly to a company's documents section.
Individual file uploads are limited to 20 MB per file, for both manual uploads and email attachments; files over 10 MB keep their extracted text only. Diligence data rooms accept files up to 100 MB. For larger files (such as high-resolution board decks or extensive spreadsheets), consider splitting them into smaller parts, compressing images, or exporting to a more compact format before uploading. The AI processing pipeline works best with focused, well-structured documents rather than very large omnibus files.
The system extracts text content from uploaded files to make them available to the AI for analysis and metric extraction. PDFs and Office documents have their text extracted server-side. Images are processed using the AI's vision capabilities to read charts, tables, and text directly from screenshots or photos of reports, and images and scanned PDF pages in company updates are queued for OCR so their text is searchable.
Every upload is checked before it is stored: blocked file types, executable content, files whose contents don't match their type, and archive bombs are rejected. This is not a full antivirus, so if your policy requires one, add scanning at the storage provider (Supabase Storage or Google Drive) as well.
When file storage is configured (Google Drive), email attachments and uploaded documents are automatically organized into company-specific folders. This provides a backup of all source materials alongside the extracted data. Files stored in Supabase Storage are accessible through the platform's UI; files in Google Drive can also be accessed directly through that service.
Uploaded files and their extracted content are only accessible to members of your fund.Row-level security limits each user to their fund and, within it, to the entities they’ve been granted, storage included. File content sent to AI providers (Anthropic, OpenAI or OpenRouter) for processing is subject to those providers' data handling policies — refer to their documentation for details on data retention and usage.
Version updates
The platform includes a built-in update checker that compares your installed version against the latest release on GitHub. When a newer version is available, admins will see an Updates link in the sidebar with an indicator dot.
The Updates page shows your current version, the latest available version, release notes, and a link to the GitHub release. The check runs automatically and is cached for one hour, so it does not slow down normal usage.
Each installation has a unique Installation ID — a UUID generated automatically in your database. This ID is specific to your deployment and is shown at the bottom of the Updates page, and is sent to GitHub with the version check. Only admins can see the Updates page; non-admin users are not shown the update indicator. (Not to be confused with Portfolio → Updates, which is company updates.)