MCP server

Your agent already reasons. Now it can test.

Point Claude, or any MCP client, at the Stax server and your agent gets the same engine this whole product runs on: it can screen thousands of companies, prove a plan on years of real prices, and put the winner in a paper slot, all in one conversation.

Get an API key Read the docs

claude· connected to stax

> find me solid large companies and put a tested plan on paper

screen_stocks{ "roe": ">15", "pe": "<20", "mcap": ">10B" }

→ 87 companies match. Top by size: AAPL, MSFT, JNJ, PG, COST…

backtest{ "symbols": ["AAPL","MSFT","JNJ","PG"], "years": 5 }

→ Done in 4.2s. 214 trades, fees included, every buy and sell listed.

deploy_strategy{ "name": "Steady large caps", "capital": 10000 }

→ Deployed to a paper slot. It now trades simulated cash you can watch.

done Your agent screened, tested, and deployed. No real cash is involved until you say so.

Connected in one command

The server speaks standard MCP over HTTP. Your key comes from the account page after a free signup.

terminal

# Add Stax to Claude Code (any MCP client works the same way)
claude mcp add --transport http stax \
  https://stax-mcp.staxvest.workers.dev/mcp \
  --header "Authorization: Bearer sk_YOUR_KEY"

# Then just talk to it:
# "screen for companies with return on equity above 15,
#  backtest the top five over five years,
#  and deploy the best one to a paper slot"

Ten tools, one honest engine

Everything your agent can do, exactly as the server names it. Screening, testing, and deploying run through the same engine the app uses, with the same no-peeking rules.

screen_stocks
Filter thousands of companies by financial health: profitability, debt, valuation, size.
backtest
Replay a plan over years of real prices, fees included, every trade recorded with its reason.
optimize
Run many backtests across parameter combinations and return the versions that held up.
list_strategies
List the plans saved on the connected account.
get_strategy
Fetch one saved plan with its full rule set.
save_strategy
Save a plan to the account so the app and the agent see the same thing.
deploy_strategy
Put a plan in a paper slot where it trades simulated cash on real prices.
list_deployments
Check every running deployment and how each is doing.
browse_community
Browse public community strategies, forkable by id.
account
Read the connected account: tier, slots in use, and limits.

claude· connected to stax

> find me solid large companies and put a tested plan on paper

screen_stocks{ "roe": ">15", "pe": "<20", "mcap": ">10B" }

→ 87 companies match. Top by size: AAPL, MSFT, JNJ, PG, COST…

backtest{ "symbols": ["AAPL","MSFT","JNJ","PG"], "years": 5 }

→ Done in 4.2s. 214 trades, fees included, every buy and sell listed.

deploy_strategy{ "name": "Steady large caps", "capital": 10000 }

→ Deployed to a paper slot. It now trades simulated cash you can watch.

done Your agent screened, tested, and deployed. No real cash is involved until you say so.

Why this is different

Your agent has to show its work.

Most AI trading demos are a model narrating its own confidence. This is the opposite: when your agent proposes a plan, the backtest tool makes it survive years of real history, with next-morning fills and fees counted, before anyone acts on it.

And because deployment goes to a paper slot, the agent can run its idea against the live market with simulated cash while you read every trade it makes. Real money stays a human decision.

Built on the same rails as the app

10
tools exposed over standard MCP: ~3s
for a thousand-run tuning pass: 1
Bearer key. No OAuth dance, no SDK required

Rate limits follow your account tier; the account tool reports yours.

Questions people ask

Which clients work with it?

Anything that speaks MCP over HTTP: Claude Code, Claude Desktop, and the growing set of agent frameworks with MCP support. If your client can add a server by URL and header, it can drive Stax.

Can my agent spend real money?

No. The MCP tools deploy to paper slots only, where plans trade simulated cash. Live brokerage connections are a separate, human-only step inside the app, on the highest tier.

Is this the same engine the app uses?

Yes, literally the same one. The MCP server is a thin layer over the Stax API, which runs the same screening and backtesting engine the product runs. Your agent and your app always agree.

What does a key cost?

Keys come with any account, including free ones. Usage follows your tier limits, and the account tool tells your agent exactly where those limits stand.