openportfolio
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@openportfoliowhat's my net worth breakdown by account?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
openportfolio
Every account as one book. Every call on the record.
Open-source, self-hosted portfolio tracker. It pulls every brokerage, pension, wallet and bank account into a single net worth, stores the investor flows behind the price, and Brier-scores the forecasts you registered before the fact. No provider API key, anywhere in it.
Screenshots show the demo book. Every figure in them is invented.
Status: pre-release. It runs and the setup below works. Interfaces will still move.
It is not a trading bot. The backend has no function that places an order, the adapters that
ship declare canPlaceOrders: false, and PlaceOrderRequest requires an OrderConfirmation that
has no default. What it does is aggregate, watch, and keep score.
Why
Two problems that look unrelated and are the same problem.
A portfolio is scattered by construction. A broker here, a pension there, an ISA, an exchange account, a bank balance, a holding that no API will ever return. Every one of those apps shows you a number, and none of them shows you yours. So the figures that actually decide things, the total, the concentration in one name, the share of the book sitting in a currency you do not spend, are the figures nobody has. They get estimated, and the estimate is generous in the direction that avoids a decision.
Market commentary is unaccountable, and became more so the moment a model would produce a confident directional view on anything you asked it. The problem is not that the views are wrong. It is that being wrong costs nothing and leaves no trace, so a forecaster worth reading and a forecaster who is merely fluent are indistinguishable from the outside, and from the inside too.
Both are bookkeeping failures, so openportfolio treats them as bookkeeping.
One net worth | Accounts pulled through venue adapters into a single base currency, with the rate stored on the row it converted, so a snapshot records what the book was worth then rather than what today's rates say. Positions held in three places are one exposure. |
Flows, not just prices | Price is the output of who was buying and who was made to sell. Net buying by investor type, turnover, and a calendar of dated forward events are stored as first-class series, not derived when somebody remembers to ask. A forced seller is on a schedule, and the schedule is public. |
A scored track record | A call is registered before the fact with a probability, a horizon and the condition that settles it. When the horizon passes, the machine-resolvable ones settle themselves and are Brier-scored. The reliability diagram is the product: it shows what you said, what happened, and the gap between them. |
There is a fourth table that exists for one reason. A recommendation phrased as "wait for the
print, then decide" evaporates the moment it is said out loud. decisions is a queue of those,
each with a trigger condition and an outcome, and they stay on the board until one of them changes.
Related MCP server: Fund Dashboard MCP Server
No provider API key
Watching a book is only useful if something is actually watching: reconciling after the close, settling a call the day its horizon passes, noticing that a deferred decision came due three weeks ago.
Metered inference is the wrong shape for that. When each run bills per token, every autonomous check becomes a purchase, and a product that spends the operator's money unprompted has to ask first, or batch, or ration. All three turn a portfolio that watches itself into a portfolio that asks permission to look.
So every model call is dispatched instead to an agent CLI you are already signed in to (codex,
antigravity, claude), with a per-task fallback order. There is no provider key in this repo and
no field to put one in. That does not make a run free: subscription plans have rate limits, and the
fallback chain exists partly because one provider runs out before the others do. What changes is
the kind of limit. Agent work is bounded by quota and wall clock rather than by spend, so it never
has to be justified one invocation at a time.
The consequence is that openportfolio is self-hosted by design. Your deployment runs your syncs on your machine under your own logins, against your own accounts.
What it does
Surface | |
Net worth | accounts, balances, per-venue and per-asset-class breakdown, snapshots in one base currency, keyless FX |
Venues | adapter contract with declared capabilities; a reference keyless quote adapter and a manual one |
Flows | net buying and turnover by investor type per session, per market or per symbol |
Forecasts | probability, horizon and resolution criterion; auto-resolution on horizon expiry; Brier score and reliability buckets |
Decisions | the deferred-decision queue, with trigger conditions and outcomes |
Catalysts | dated forward events and the assets they touch |
Audit | append-only record of every state-changing mutation, including what the cron did unattended |
MCP | 25 tools so |
Multi-tenancy | every table scoped to a tenant, every index leading with it, one service key per tenant |
Quick start
Node 22+, pnpm, and a Convex account. The free tier is enough.
git clone https://github.com/seonglae/openportfolio.git
cd openportfolio
pnpm install
cp .env.example .env.local
npx convex dev --once # creates the deployment
# create the first book
npx convex env set OPENPORTFOLIO_DEV_TENANT home
npx convex run tenants:create '{"slug":"home","name":"Home","baseCurrency":"GBP"}'
# the UI, then the sync loop
pnpm --filter openportfolio-browser dev # http://localhost:6101
npx tsx sync-worker.mts --onceWith nothing linked it registers the venues it can serve and records a net worth of zero, which is correct. Add a manual holdings file to get a real one:
[
{ "accountKey": "isa", "symbol": "VWRL", "assetClass": "etf", "qty": 40, "price": 118.2, "currency": "GBP" },
{ "accountKey": "wallet", "symbol": "BTC", "assetClass": "crypto", "qty": 0.15, "price": 0, "currency": "USD" }
]export OPENPORTFOLIO_MANUAL_HOLDINGS=$PWD/holdings.json
npx convex run accounts:link '{"accountKey":"isa","venue":"manual","kind":"brokerage","label":"ISA","currency":"GBP"}'
npx convex run accounts:link '{"accountKey":"wallet","venue":"manual","kind":"wallet","label":"Wallet","currency":"USD"}'
npx tsx sync-worker.mts --onceThe BTC row is priced at 0 in the file on purpose: the worker re-quotes crypto through the keyless CoinGecko adapter, converts both rows into GBP, and writes one total.
Full walkthrough: openportfolio.app/docs/quickstart
Before exposing it
Two things are open on localhost and must be closed before the deployment is reachable from the internet.
The dev tenant. While
OPENPORTFOLIO_DEV_TENANTis set, any unauthenticated caller is scoped to that tenant. Unset it and configure Clerk.Service keys. Workers and the MCP server have no browser session, so they present a key. Generate it locally and send only its hash.
npx convex env set CLERK_ISSUER_URL https://your-app.clerk.accounts.dev
npx convex env unset OPENPORTFOLIO_DEV_TENANT
KEY="$(openssl rand -hex 32)"
npx convex run tenants:issueServiceKey "{\"key\":\"$KEY\",\"label\":\"sync-worker\",\"role\":\"member\"}"
echo "OPENPORTFOLIO_SERVICE_KEY=$KEY" >> .env.localMulti-tenancy
One deployment holds many books. The invariant is that a caller never says which tenant it is.
tenantId is derived from the caller's membership rows or from the service key's own row, so there
is no argument a client can set to reach another book. The public API accepts tenantSlug, and
only as a disambiguator for a caller who belongs to several tenants; membership is still what
decides. A document id belonging to another tenant reads as missing rather than forbidden, because
"forbidden" confirms the row exists, which is itself the cross-tenant read.
Every index leads with tenantId, so a query that forgets the scope cannot use an index at all.
One exception is deliberate and marked: the resolver cron sweeps every book's due calls through a
tenant-less index, and is an internalMutation for exactly that reason. It is unreachable from any
client.
Details: openportfolio.app/docs/multi-tenancy
Venue adapters
An adapter declares what it can do and implements only that:
type VenueAdapter = {
venue: string;
kind: AccountKind;
capabilities: { canReadBalances: boolean; canReadQuotes: boolean; canPlaceOrders: boolean };
readBalances(request: ReadBalancesRequest): Promise<AdapterBalance[]>;
readQuote(request: ReadQuoteRequest): Promise<AdapterQuote>;
placeOrder?(request: PlaceOrderRequest): Promise<OrderReceipt>;
};Two ship. coingecko reads quotes and refuses balances, because a price source does not know what
you hold and returning an empty list would read as "you hold nothing". manual reads a JSON file
you maintain, which is how a pension or an unlisted holding gets into the total instead of being
left out of it.
No keyed broker adapter ships. Adding one means writing a module in
packages/node/src/adapters/, taking its credential from the worker's environment, and registering
it in defaultRegistry(). Keep the credential in the worker process: the backend never sees it,
and neither does this repo.
Details: openportfolio.app/docs/adapters
Requirements
Node 22+, pnpm
A Convex account (free tier is enough)
At least one agent CLI signed in, if you want the agent worker:
codex,antigravity(agy), orclaudeOptional: Clerk for auth, needed once more than one person uses the deployment or it is reachable from the internet
Development
pnpm typecheck # every workspace, src and test alike
pnpm test # vitest across packages, convex handlers, browser helpers
# the demo build used for the screenshots and the hosted demo
pnpm --filter openportfolio-browser exec vite build --config vite.demo.config.ts
# the marketing site and docs are static; regenerate the docs pages after editing
python3 site/build-docs.pyConventions, the tenant invariant in full, and notes for agent CLIs working in this repo are in AGENTS.md.
License
Apache-2.0. See LICENSE.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- Flicense-qualityDmaintenanceEnables AI agents to interact with the WYGIWYH expense tracking API through 75 dynamically generated MCP tools. Supports comprehensive financial operations including transaction management, account handling, recurring expenses, and investment tracking.7
- Alicense-qualityBmaintenanceEnables AI agents to manage and analyze personal investment portfolios, including fund and stock holdings, net value tracking, XIRR calculations, penetration analysis, and backtesting.Apache 2.0
- Flicense-qualityCmaintenanceEnables AI agents to manage personal finances through MCP tools for transaction management, spending analytics, and goal tracking.1
- Flicense-qualityBmaintenanceEnables AI agents to manage personal finances for Brazilian users through MCP tools, including categorizing transactions, reconciling debts, checking cash-flow projections, and adjusting budgets, with integration to Open Finance Brasil via Pluggy.
Related MCP Connectors
Real SEC, 13F, insider, congress & macro data your AI agent can cite. Hosted MCP, 24 tools.
The financial MCP for AI agents - 90+ financial tables, SEC filings, signals, alt-data.
Shared long-term memory vault for AI agents with 20 MCP tools.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/seonglae/openportfolio'
If you have feedback or need assistance with the MCP directory API, please join our Discord server