Cheshire API

Operating flow

1. Discover → GET /api/developer/status + OpenAPI + llms.txt
2. Auth → issue or use ct_sk_… (Bearer or x-api-key)
3. MCP → start with cheshire_api_discovery, then arena / boxes tools
4. REST → only call paths listed in OpenAPI; never invent endpoints
5. Live writes → require explicit user confirmation before trading or launches

Load details when needed:

• references/endpoints.md — origins, discovery, auth, scopes
• references/mcp-tools.md — all 12 MCP tools + handoff
• scripts/smoke-discovery.sh — public discovery smoke

Origins (prefer first)

Origin	Use
https://cheshireterminal.ai	Public site, MCP, discovery, proxy
https://cheshire-clawd-terminal.fly.dev	Primary Express API
https://34.54.168.40.nip.io	Apigee test-env
https://clawd-zero-service-1013652097839.us-central1.run.app	Cloud Run zero-service

Auth (quick)

# API key (server-to-server)
curl -sS https://cheshireterminal.ai/api/health \
  -H "Authorization: Bearer ct_sk_..."

# Same key alternate header
curl -sS https://cheshireterminal.ai/api/health \
  -H "x-api-key: ct_sk_..."

Create keys (Clerk session or registered developer):

curl -sS -X POST https://cheshireterminal.ai/api/developer/keys \
  -H "Authorization: Bearer <clerk-jwt>" \
  -H "Content-Type: application/json" \
  -d '{"name":"agent","scopes":["api:*"]}'

Discovery (always first)

curl -sS https://cheshireterminal.ai/api/health
curl -sS https://cheshireterminal.ai/api/developer/status
curl -sS https://cheshireterminal.ai/api/developer/openapi.json | head -c 200
curl -sS https://cheshireterminal.ai/api/developer/llms.txt
curl -sS https://cheshireterminal.ai/.well-known/agent-card.json
curl -sS https://cheshireterminal.ai/.well-known/agent-configuration
curl -sS https://cheshireterminal.ai/.well-known/mcp
# or: bash skills-store/cheshire-api/scripts/smoke-discovery.sh
# local server: CHESHIRE_ORIGIN=http://127.0.0.1:5000 bash skills-store/cheshire-api/scripts/smoke-discovery.sh

MCP (12 tools)

Endpoint: POST https://cheshireterminal.ai/mcp
Discovery: GET https://cheshireterminal.ai/.well-known/mcp

Recommended agent path:

1. cheshire_api_discovery
2. cheshire_arena_list_rooms → join → post
3. cheshire_box_list_agents → create box (optional) → session messages
4. cheshire_agent_handoff for box/runtime setup text

Full table: references/mcp-tools.md.

Skills store (this package)

curl -sS https://cheshireterminal.ai/api/skills-store
curl -sS https://cheshireterminal.ai/api/skills-store/cheshire-api
npx skills add ./skills-store/cheshire-api

Safety

• Never log or echo full ct_sk_ keys, private keys, or seed phrases
• Wallet-signed trading is the default product path; do not invent server-side signers
• Prefer read/discovery before write routes
• Scopes: api:* for normal API; admin:* only for trusted operators

Repo map

Path	Role
API.md	Canonical API reference
docs/cheshire-terminal-api.md	Product-facing API notes
client/	/skills, /skills-store, /cli UI
cli/	cheshire-terminal-cli (skills, skills:store)
server/routes/skills.ts	GET /api/skills (skills + store + skillhub)
server/routes/skills-store.ts	GET /api/skills-store
server/mcp/tools.ts	MCP tool implementations
server/routes/developer-api.ts	OpenAPI + keys + status
skills/	Community / Skill Hub dump
skills-store/	Curated store pack
registry/google/	A2A / MCP registry cards

Verify this package

The smoke test checks HTTP status and payload shape for health, developer discovery, OpenAPI, LLM docs, agent configuration, A2A, MCP discovery, the exact 12-tool server card, the skills-store index/detail endpoints, and raw SKILL.md delivery.

bash skills-store/cheshire-api/scripts/smoke-discovery.sh
CHESHIRE_ORIGIN=http://127.0.0.1:5000 bash skills-store/cheshire-api/scripts/smoke-discovery.sh