OpenAlice Trading MCP
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., "@OpenAlice Trading MCPstage a buy of 10 shares of AAPL"
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.
OpenAlice Trading MCP
Standalone MCP server exposing OpenAlice's Trade-as-Git trading workflow
to any MCP-capable AI agent (OpenClaw, Claude Desktop, Cursor, ...). The
trading domain — broker connections, the git-like approval state machine
(stage / commit / push / reject), FX, snapshots, every IBroker engine
(CCXT, Alpaca, IBKR TWS port, Mock, Leverup) — is the original services/uta
service from OpenAlice, lifted out of the Alice + Dashboard runtime and
rewired around the agent-facing surface only.
智能自动交易: 通过 OpenClaw HEARTBEAT 机制实现自动监控和执行。用户只需要用自然语言描述策略(如"帮我监控 AAPL,止损 170"),AI 会自动配置监控并在触发条件时执行交易。支持增量更新规则、调整检查频率、停止监控等操作。
There is no Web UI in this repo. The Trade-as-Git approval gate that the
OpenAlice Dashboard provides is replaced by a skill (skills/openalice-trading/SKILL.md)
that teaches the consuming agent the user-approval ritual: stage and commit
are AI-free, but tradingPush requires explicit user consent obtained in
chat. v1 trusts the skill; the server does not enforce a server-side
approval gate.
Architecture
┌─ apps/mcp-server (this repo) ───────────────────────────────┐
│ │
│ ┌─ supervisor ─────────────────┐ │
│ │ spawn services/uta child │ internal: 127.0.0.1:47333│
│ │ wait for /__uta/health │ │
│ │ SIGTERM cascade on shutdown │ │
│ └──────────────────────────────┘ │
│ │
│ ┌─ MCP Streamable HTTP ────────┐ │
│ │ register 19 trading tools │ exposed: 127.0.0.1:47400 │
│ │ wraps Vercel-ai SDK shape │ path: /mcp │
│ │ via @openalice-trading/ │ │
│ │ trading-tools │ │
│ └──────────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘
▲
│ MCP protocol (HTTP+SSE)
│
┌──────────────────┴───────────────────┐
│ OpenClaw (or any MCP client) │
│ + skills/openalice-trading/SKILL.md │
└───────────────────────────────────────┘Related MCP server: eToro MCP Server
Layout
.
├── packages/
│ ├── alice-core/ # minimal subset of OpenAlice's core extracted
│ │ # for UTA's startup path (config schemas,
│ │ # event log, paths, sealing, market-data
│ │ # client ports, mcp-export bridge)
│ ├── ibkr/ # @traderalice/ibkr — TS port of IBKR TWS API
│ ├── opentypebb/ # @traderalice/opentypebb — TS port of OpenBB
│ ├── trading-tools/ # the 19 AI tools + UTAManagerSDK / UTAAccountSDK
│ └── uta-protocol/ # wire types + zod + UTA HTTP client
├── services/
│ └── uta/ # the unmodified UTA service (broker engines,
│ # TradingGit, FX, snapshots) — runs as a child
│ # process of mcp-server
├── apps/
│ └── mcp-server/ # the user-facing entry: supervises UTA +
│ # exposes trading tools over Streamable HTTP MCP
└── skills/
└── openalice-trading/
└── SKILL.md # the user-approval ritual the AI must followQuick start
pnpm install # ~30s, includes broker SDKs
# (one-time) seed broker accounts. Either:
# (a) reuse OpenAlice's: ln -s ~/.openalice/data ~/.openalice-trading-mcp/data
# and ln -s ~/.openalice/sealing.key ~/.openalice-trading-mcp/sealing.key
# (b) start fresh — the server will create an empty data store.
OPENALICE_HOME=$HOME/.openalice-trading-mcp \
pnpm -F @openalice-trading/mcp-server devThe server logs:
[mcp] bootstrap @ 2026-06-19T...
[mcp] OPENALICE_HOME = /home/.../.openalice-trading-mcp
[mcp] AGENT_AUTHOR = openclaw
[mcp] spawning UTA child: ...
[uta] [uta] listening on http://127.0.0.1:47333
[mcp] UTA ready ... (N accounts)
[mcp] registered 19 trading tools
[mcp] listening on http://127.0.0.1:47400/mcp (Streamable HTTP)Health check: curl http://127.0.0.1:47400/
OpenClaw client config
Per OpenClaw MCP docs, register the
server with transport: "streamable-http":
openclaw mcp set openalice-trading '{
"transport": "streamable-http",
"url": "http://127.0.0.1:47400/mcp",
"connectionTimeoutMs": 5000
}'Install the skill:
mkdir -p ~/.openclaw/skills
cp skills/openalice-trading/SKILL.md ~/.openclaw/skills/openalice-trading.md(Adjust to OpenClaw's actual skill directory; the docs are the authority.)
Configuration
Env var | Default | Purpose |
|
| User data root. Holds |
|
| Where the MCP server listens. |
|
| Internal UTA child port (loopback only). |
|
| Boot deadline for the UTA child to become healthy. |
|
| Tag written into TradingGit commits. Helps disambiguate when multiple agents share one UTA (OpenAlice = |
Smoke test
A scripted end-to-end smoke is included:
# Server up in one terminal:
OPENALICE_HOME=/tmp/test-trading-home pnpm -F @openalice-trading/mcp-server dev
# In another:
mkdir -p /tmp/test-trading-home/data/config
cat > /tmp/test-trading-home/data/config/accounts.json <<'EOF'
[{
"id": "mock-smoke",
"label": "Smoke Mock",
"presetId": "mock-simulator",
"enabled": true,
"guards": [],
"presetConfig": { "cash": 100000, "_instanceId": "smoke-test-001" }
}]
EOF
# (restart server to pick up the account)
cd apps/mcp-server
node --conditions=openalice-source --import tsx scripts/smoke.tsThe script drives placeOrder → commit → tradingPush → portfolio → closePosition → push → reject against the in-memory mock broker and prints
each tool call's result.
Production Deployment (systemd)
For production use, set up a systemd service to manage the MCP server:
# Copy the service file
cp trading-mcp.service ~/.config/systemd/user/
# Reload systemd and enable the service
systemctl --user daemon-reload
systemctl --user enable trading-mcp
# Start the service
systemctl --user start trading-mcp
# Check status
systemctl --user status trading-mcp
# View logs
journalctl --user -u trading-mcp -fThe service file is configured to:
Start automatically on boot
Restart on failure (with 10s delay)
Run as your user (not root)
Use
/tmp/openclaw-testas the data directoryListen on port 47400
To stop or disable:
systemctl --user stop trading-mcp
systemctl --user disable trading-mcpWhat's NOT in this repo
No Web UI / Dashboard. (Skill replaces the approval gate.)
No market-data / news / analysis tools. (Out of scope; OpenClaw's built-in capabilities cover those.)
No Alice-side workspace launcher. (That's OpenAlice's job, not this server's.)
No telegram / mcpAsk connectors. (Those are upstream concerns.)
Extending
New tools: add to
packages/trading-tools/src/tools.ts(Vercel-ai SDK shape), they will auto-register on next start.New brokers: same path as OpenAlice — add an
IBrokerimpl underservices/uta/src/domain/trading/brokers/<name>/and register it inservices/uta/src/domain/trading/brokers/registry.ts.Stricter approval: if you ever want server-side enforcement (don't trust the skill), wrap
tradingPush.executeintrading-tools/tools.tswith a token-handshake — push requires aapproval_tokenreturned by a priortradingStatuscall. See the Anti-patterns section ofSKILL.mdfor the rationale either way.
License
AGPL-3.0-only (inherits from OpenAlice).
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.
Latest Blog Posts
- 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/orangeZSCB/trader-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server