tdavidson/portfolio
tdavidson/portfolio has 75 GitHub stars and 27 forks. GitHot measured 0 stars in 24 hours, +1 in 7 days, and +1 in 15 days. Last checked 2026-11-17 03:25 UTC. Last push 2026-11-06.
What you can do
- Data consistency and availability - One source of truth for your team. Reduce your reliance on a maze of spreadsheets. Everyone works from the same portfolio data, metrics, and reports from a central location.
- Built to work with AI - Bring your fund data to your own AI, and use it to ask anything about your portfolio and fund. Ask about benchmarks, trends, industry data, research, and more.
- Professionalize internal operations - Institutional-quality reporting infrastructure without the cost of enterprise software. Run it yourself, on your own terms.
- Built for how funds work - Designed by a fund CFO for key workflows, including investor updates, LP reporting, and portfolio monitoring. Works alongside your fund admin and operations team.
Measured star growth
- in 24h
- 0
- in 7 days
- +1
- in 15 days
- +1
- in 1 month
- Collecting history
Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.
About
Manage deal flow, portfolio reporting, fund operations, LPs, and accounting, without stitching together spreadsheets and multiple tools. Open source, self-host, powered by your AI.
Self-hostedREADME.md
Original
Portfolio, by Hemrock
Manage deal flow, portfolio reporting, fund operations, LPs, and accounting, without stitching together spreadsheets and point tools.
Portfolio is an open-source operating platform for venture capital firms. Take a company from inbound screening and diligence through investment, portfolio monitoring, fund accounting, and LP reporting—all in one system. Turn on only the workflows you need, deploy on your own infrastructure, and use your own AI providers.

What it does
Portfolio gives your team one source of truth from the first look at a deal through the life of the investment. Forward founder updates, import existing data, and let AI extract the metrics, investments, and cash flows that keep portfolio dashboards and reports current.
The investment workflow classifies and fit-scores inbound opportunities, runs your diligence checklist against the data room, verifies claims with sourced external research, and drafts investment memos in your firm's structure and voice.
Fund operations run on real double-entry books that keep themselves: every entity has its ledger from the day it is created, and recording an investment, a mark, an exit or a capital call posts it — so capital accounts, statements, carry and LP reporting flow from the books whether or not anyone opens the accounting pages. Turn on the views you need, when you need them. A secure LP portal delivers statements, letters, and fund documents and records engagement.
How it works
- Portfolio Reporting — Forward investor updates in any format or import existing spreadsheets. AI identifies each company, extracts metrics, investments, and cash flows, and sends uncertain results to a review queue. Track custom KPIs, valuations, proceeds, TVPI, DPI, and Net IRR in current dashboards.
- Investment Workflow — Classify and thesis-score inbound pitches, manage the deal pipeline, run your diligence checklist against uploaded data rooms, verify claims through sourced external research, and draft an investment memo in your firm's voice with paragraph-level provenance.
- LP Reporting — Track commitments, paid-in capital, distributions, and NAV across vehicles as dated positions; generate statements and quarterly letters; and deliver reports and documents through a secure, fund-branded LP portal with engagement tracking. Investors can also connect Claude or ChatGPT to their own portal.
- Fund Operations — Double-entry books for every entity, kept as you record investments and capital calls; import bank activity, run monthly closes, maintain partner capital accounts, model management fees and carried interest, budget and forecast each entity, and track compliance obligations.
- Your AI, your data — Connect Claude or ChatGPT to your own deployment over MCP with a plugin built for your address. Ask "what can you do?" for a home dashboard of what you can open and what to ask; open live portfolio, company, statement, LP capital and capital-call dashboards inside the conversation (who has paid, who says they wired, who still owes) and save them; pick ready-made prompts like Quarter-end review from the prompt menu; and get an LP's capital account statement or report card as a PDF link — every answer scoped to the access of the person signed in. No server AI key is required: without one, your own assistant does the AI work, from reading emailed updates and scanned pages to running the memo agent's stages, and the web app's AI buttons are the main thing you give up (what works without a key).
- Headless by design — Chat and scripts are first-class ways in and out, not an add-on to the web app. Set up a fund, add the team, record investments, post entries and close periods from Claude, ChatGPT or your own agent (MCP and a REST API); read anything back the same way. The web app is a view of the same data, and the preformatted PDFs and reports stay. Every change, whichever way it comes in, passes the same access rules and is recorded in the audit trail with its source (web, API or MCP); a change from a connected assistant waits for a person to approve it. Data from other systems comes in through private connectors you keep in your own repository (a template for QuickBooks- and Carta-style ledgers is in
examples/connector-template). - AI Analyst and CRM — Ask questions grounded in your fund's actual data, compare companies, surface trends, share Notes with your team, and log conversations and introductions through your inbound address.
- Access by entity and by area — Grant each team member the funds, SPVs and entities they work on, and the areas (portfolio, accounting, LP capital, carry, tax…) they may read or edit, in one step with presets for a CFO, fund administrator, tax preparer, auditor, investment team or investor relations, and with an end date for anyone here for one engagement. Deletes, moving money, anything sent to LPs and finalizing K-1s need an admin. Enforced the same way in the app, over the API and over MCP, with entity scoping also enforced in the database.
- Modular and self-hosted — Enable only the workflows you use, add all of your funds, SPVs, and team members without per-seat fees, and install the fund-branded app on a phone.
Detailed feature descriptions at FEATURES
Fund accounting, forecasting and the double-entry reference at ACCOUNTING
Connecting Claude and ChatGPT at DOCS
What the app can do that MCP and the API can't yet, and why, at PARITY
Setting up a fund from chat or a script, with no web app at DOCS
Bringing data in from QuickBooks, Carta and other systems with connectors at DOCS

Why you should use this
- Data consistency and availability - One source of truth for your team. Reduce your reliance on a maze of spreadsheets. Everyone works from the same portfolio data, metrics, and reports from a central location.
- Built to work with AI - Bring your fund data to your own AI, and use it to ask anything about your portfolio and fund. Ask about benchmarks, trends, industry data, research, and more.
- Professionalize internal operations - Institutional-quality reporting infrastructure without the cost of enterprise software. Run it yourself, on your own terms.
- Built for how funds work - Designed by a fund CFO for key workflows, including investor updates, LP reporting, and portfolio monitoring. Works alongside your fund admin and operations team.
Why this exists
I've spent over a decade as a fund CFO, investor, and consultant — working with thousands of GPs and founders on the exact problem this tool solves: manually collecting, analyzing, and presenting portfolio data every quarter.
Most portfolio reporting platforms lock your data in their database, process it through their AI, and charge per seat so half your team can't log in. Fund managers shouldn't have to choose between good tooling and owning their data.
This is a complete investment-firm operations platform you deploy on your own infrastructure — your database, your AI keys, your domain. It's open source under the Apache 2.0 license: free to use, modify, and run forever, for your own fund or commercially. No per-seat fees. No black-box AI training on your portfolio. No vendor lock-in.
Built by Taylor Davidson at Hemrock. Built by a fund manager, for fund managers. Product site and documentation: hemrock.com/portfolio.
Get started
Free and open source under the Apache 2.0 license — use it, modify it, and deploy it on your own infrastructure and domain, for your own fund or commercially. Try the demo with sample data, no signup required.
Prefer not to run it yourself? Taylor offers paid setup & support: deployed on your own infrastructure and accounts, with onboarding and ongoing support. Contact Taylor to discuss.
See LICENSE for full terms.
Quick start
- Clone the repo —
git clone https://github.com/tdavidson/portfolio.git && cd portfolio && npm install - Create a Supabase project — Copy your project URL, anon key, and service role key
- Generate an encryption key — openssl rand -hex 32
- Deploy to Netlify or Vercel — One-click deploy buttons available in the full guide
- Configure auth and add your first user — Set Supabase redirect URLs and whitelist your email
- Add an AI key, or skip it and use your own assistant, and forward your first email — Anthropic, OpenAI or OpenRouter on the server, or Claude or ChatGPT over MCP with no key: AI without a server key
- Set up your fund — entity, LPs, investments, then open the books, with CSV templates: Setting up a new fund
Or let a script do most of it: after cloning, npm run setup creates the database (with the Supabase CLI), deploys (with the Vercel CLI), configures sign-in and creates your fund, offering to install either CLI you don't have and printing the manual step for anything it can't do. It also asks how AI should work: with a server key, or with no key, where your own Claude or ChatGPT does the AI work over MCP. See The quick way.
Or set it up headless. After the database and deploy steps, npm run bootstrap -- --email [email protected] --fund "Your Fund" creates your account, the fund and an API key without opening the web app, or add your email to allowed_signups, connect Claude or ChatGPT to https://<your-domain>/api/mcp and name your fund on the consent screen. Then run the Set up this fund prompt: Setting up a fund with Claude, ChatGPT or a script
Full deployment guide with detailed steps, optional services, and local development setup: DOCS
For setup assistance or hosted deployments: hemrock.com/contact. For bug reports and feature requests: GitHub Issues.