openInvest — Claude Code MCP server: what it does, how to install
openInvest gives AI agents an investment committee, portfolio ledger, and traceable decision records. Use it for research, backtests, and review.
claude mcp add --env INVEST_HOME=<INVEST_HOME> --transport stdio openinvest -- uvx openinvestWhat openInvest does
openInvest is a self-hosted decision engine for AI agents. It runs a multi-agent investment committee with separated analysis contexts and rebuttal rounds. A coordinator assigns the work, while the recorded trail keeps the reasoning behind each decision visible.
The server maintains a portfolio ledger, supports long-horizon backtesting, and compares results with control benchmarks. Once self-hosted, its chart reads your own memory/portfolio_history.jsonl and appears after the ledger covers 2 trading days. The project lists Python 3.13+ for the runtime. Its own audit positions the system as a way to enforce discipline and transparency, not as a source of excess returns or a directional prediction engine.
What it helps with
- run an investment committee with separate analyst roles
- keep quant and risk contexts apart
- record why a portfolio decision was made
- compare a portfolio with control benchmarks
- test a trading idea with long-horizon backtesting
Example prompts for the agent
- Ask the investment committee to assess an asset, challenge the first view, and record each role’s reasoning.
- Run a long-horizon backtest for this strategy and compare the result with the control benchmarks.
- Review the portfolio decision trail and explain the reasoning behind the latest action.
How to install openInvest
Platform CLI
claude mcp add --env INVEST_HOME=<INVEST_HOME> --transport stdio openinvest -- uvx openinvestReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from openinvest <command>
Platform CLI
codex mcp add openinvest --env INVEST_HOME=<INVEST_HOME> -- uvx openinvestReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from openinvest <command>
Config file
{
"mcpServers": {
"openinvest": {
"command": "uvx",
"args": [
"openinvest"
],
"env": {
"INVEST_HOME": "<INVEST_HOME>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from openinvest <command>
Config file
{
"servers": {
"openinvest": {
"type": "stdio",
"command": "uvx",
"args": [
"openinvest"
],
"env": {
"INVEST_HOME": "<INVEST_HOME>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from openinvest <command>
Platform CLI
gemini mcp add -e INVEST_HOME=<INVEST_HOME> openinvest uvx openinvestReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from openinvest <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about openInvest
How do I install openInvest in Claude Code?
Run claude mcp add --env INVEST_HOME=<INVEST_HOME> --transport stdio openinvest -- uvx openinvest.
What does openInvest do?
It adds a multi-agent investment committee, isolated analysis, rebuttal rounds, portfolio records, and an auditable trail to an AI agent.
What does openInvest require to run?
The project lists Python 3.13+. Its self-hosted chart uses memory/portfolio_history.jsonl and appears after the ledger spans 2 trading days.
