Skip to main content
Abacus Labs is an API and MCP server for agents that help people with their investment accounts. Your agent sends an account (holdings and tax lots from any broker) and a question or a goal in plain English. Abacus Labs answers with numbers, and with lists of broker-ready trades that are checked against the account’s rules, taxes and wash sales.
Abacus Labs does not hold assets, place orders, forecast returns or give investment advice. It returns options and their consequences. The person decides. Your agent places the orders at the broker.

Start in one call

The answer has a sandbox key (abacus_sk_test_...) and three demo accounts. No sign-up. Then follow the Quickstart.

The calls

Base URL: https://api.abacuslabs.ai/v1. Send Authorization: Bearer <key> on every call. Primitives are plain reads with no AI step: Also: POST /v1/keys (sandbox key, no auth), POST /v1/certificates/verify and GET /.well-known/jwks.json (check a commit’s signature, no auth), and GET /agent.md, GET /llms.txt, GET /mcp.json (onboarding files for agents).

The usual flow

  1. Analyze the account the person has.
  2. Propose options for what the person wants. Show them all of them.
  3. The person picks one. Check its trades right before you place them.
  4. Place the orders at the broker, with the tax lots each sell names.
  5. Commit the option, so later checks know about its loss sales.

Quickstart

The flow above on a demo account, with real responses.

Keys and limits

Sandbox and live keys, claiming a key, rate limits.

MCP server

The same calls as MCP tools, for Claude Code, Cursor and your own agent.

Credits and plans

What each call costs, effort, max_credits, and 402 errors.