Limowo / Agents
Your agent can help.You stay in control.
Connect Codex, Claude, ChatGPT, or another compatible agent to search Limowo, prepare a listing, and communicate with confidence. Every real-world action returns to you.
Human-owned
Bounded actions
Swiss marketplace
Agents prepare. Owners review and decide.
Local tools. Clear checkpoints.
Browse marketplaceSearch marketplace
Public agents can discover active listings and categories without credentials through typed read commands.
marketplace.articles.searchPrepare listing draft
Authorized agents can prepare draft data for the represented owner. Owner, author, and sender identities are derived from the grant.
listing.draft.createReview with owner
Draft creation returns safe next actions so the human owner can review, edit, approve, and publish from Limowo.
listing.draft.validateRoute safely
Commands return route hints only for public listing/search pages or authenticated owner review pages.
nextActions[]Bring Limowo into your workflow.
Start with a compatible client. Public discovery works without authentication; write actions remain permissioned and human-led.
Connect a client.
Use the official Limowo connection guide to link ChatGPT, Codex, Claude, or another compatible client through OAuth. Limowo tokens and shared secrets stay out of your hands.
Open connection guideHuman-owned by design
Agents can suggest and prepare. You decide what happens next.
Action boundaries
No publishing, payments, or account changes happen autonomously.
Built in Switzerland
A local marketplace with clear permissions and accountable tools.
Everything you need to inspect the surface.
/fr/agentsHuman-readable portal for safe web agent workflows.
/llms.txtPlain-text discovery summary for assistants and indexers.
/.well-known/limowo-agent.jsonCompact machine identity and rollout metadata.
/mcp-server.jsonOfficial remote-server metadata prepared for MCP Registry publication.
/mcpProvider-neutral Streamable HTTP endpoint for MCP clients.
/api/agent/v1/openapi.jsonVersioned contract generated from the active command registry.
Safety boundaries
- 01No arbitrary shell, SQL, GraphQL documents, filesystem operations, or unrestricted HTTP.
- 02No autonomous article publication, offers, orders, payments, or account administration in V1.
- 03Write commands require a Keycloak MCP token and idempotency; publication stays with the human owner.