Archer MCP API

Archer's trading, research, and charting tools are exposed via the Model Context Protocol — the open standard for connecting AI agents to tools. Point any MCP-capable client (Claude, an SDK agent, your own loop) at the endpoints below.

Base URL

Authentication

• Every request needs Authorization: Bearer <token>. Sign in above to get your personal key instantly; partner-grade keys with wider grants come from the team (HELP in any Archer chat).
• Your token is bound server-side to your Archer identity. You never send user identifiers; any platform / platform_id fields in your arguments are overwritten at the gateway, so a token cannot act as anyone but its own identity.
• Treat the token like a private key: server-side only, never in browser code or repos. Rotation is immediate on request.

# Every request: Bearer token + both Accept types.
curl -X POST https://archer.tristero.com/mcp/hyperliquid \
  -H "Authorization: Bearer $ARCHER_MCP_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Response — note the SSE framing (event:/data: lines); parse the JSON after data:. Most MCP SDKs handle this for you:

HTTP/2 200
content-type: text/event-stream

event: message
data: {"result":{"tools":[
  {"name":"run_tool",  "description":"Invoke any cataloged tool by name …"},
  {"name":"list_tools","description":"Full searchable catalog for this server …"}
]},"jsonrpc":"2.0","id":1}

Handshake

Standard MCP initialize first if your client does not send it automatically:

{"jsonrpc":"2.0","id":0,"method":"initialize","params":{
  "protocolVersion":"2025-03-26","capabilities":{},
  "clientInfo":{"name":"my-agent","version":"1.0"}}}

// response:
event: message
data: {"result":{"protocolVersion":"2025-03-26",
  "capabilities":{"tools":{}},
  "serverInfo":{"name":"proxy-hyperliquid","version":"0.0.1"}},
  "jsonrpc":"2.0","id":0}

Calling tools

Each server exposes two meta-tools — list_tools (a compact catalog; pass names for selected schemas or detail: "full" for an export) and run_tool (invoke by name) — so your agent's tool budget stays flat regardless of catalog size. Tool descriptions carry usage guidance your model should read.

{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{
  "name":"run_tool",
  "arguments":{"tool":"hl_get_price","arguments":{"coin":"BTC"}}}}

// response (SSE-framed, real exchange):
event: message
data: {"result":{"content":[{"type":"text","text":
  "{\n  \"coin\": \"BTC\",\n  \"price\": \"64852.0\",
   \n  \"change_24h\": \"0.85%\",
   \n  \"funding_rate\": \"-0.00047%\", … }"
  }]},"jsonrpc":"2.0","id":2}

Charts

Chart-producing tools return a chart_url; GET it for a finished PNG. Candles + volume + MAs, RSI/MACD/Bollinger panels, multi-symbol comparisons, funding-rate history, order-book depth, PnL curves — all server-rendered in Archer's house style.

{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{
  "name":"run_tool",
  "arguments":{"tool":"hl_price_chart","arguments":{
    "coin":"BTC","interval":"1d",
    "indicators":["rsi","macd"],
    "compare":["ETH"],"compare_style":"percent"}}}}

// result payload (inside content[0].text):
{"coins":["BTC","ETH"], "chart_type":"compare",
 "chart_url":"https://…/render?coins=BTC%2CETH&…",
 "scale_note":"Comparison chart plots PERCENT CHANGE …"}
// GET chart_url → finished PNG.

Errors

// Tool-level failures come back HTTP 200 with isError:true —
// inspect content[0].text, do not retry blindly:
event: message
data: {"result":{"content":[{"type":"text","text":
  "Error: … guidance on whether/how to retry …"}],
  "isError":true},"jsonrpc":"2.0","id":2}

Money-path semantics — read this one

• Order tools queue a signed intent; nothing executes until the account owner confirms in their own Archer chat. A successful order call means queued for confirmation, not filled.
• Fills return receipts with on-chain transaction links.
• Sub-minimum orders are auto-sized to the venue minimum and say so in the result.

Servers

Default grants cover the servers below; additional surfaces (wallet, swaps, prediction markets) are enabled per arrangement.

Good-citizen notes

• Upstream venues rate-limit; batch reads and cache market metadata.
• Tool results may embed guidance notes (e.g. how to describe a comparison chart's scale). They exist because they prevent real mistakes — surface them to your model.
• The catalog evolves; re-run list_tools rather than pinning names forever.