API & agents

Everything in the console and CLI is a JSON API, designed to be driven by AI agents as much as humans.

Machine-readable everywhere


`/llms.txt`	the platform, summarized for LLMs
`/openapi.json`	full OpenAPI 3 spec
`/AGENTS.md`	conventions to drop into any repo
`wess <cmd> --json`	every CLI command emits JSON

Skills — teach your agent the platform

Six focused skills, served by the platform itself so they always match the deployed version (index): wess-deploy, wess-debug, wess-db, wess-ai, wess-sandbox, wess-static.

terminal

wess skill install                          # all of them, into .claude/skills/
wess skill install wess-deploy,wess-db      # a selection
wess skill install --token read,deploy,logs # + mint a least-privilege token
# no CLI? curl -fsSL https://wess.dev/skills/install | sh

--token prints a scoped WESS_TOKEN so the agent never holds your session.

MCP server

For MCP-speaking agents, the CLI doubles as a stdio MCP server exposing apps, deploys, logs, doctor, secrets, and the sandbox as tools (your token's scopes still apply):

terminal

claude mcp add wess -- wess mcp

Auth

POST /api/login for a session token, or create a scoped token for automation. Send it as Authorization: Bearer <token>.

Errors teach

Every error returns {"error": "..."} with a message written so an agent (or human) can recover — what was wrong and what to do about it.