Ontoloop

Ontoloop MCP Server

Ontoloop is a team agent collaboration platform. It connects everyone's Claude Code, Codex and other MCP clients to one shared team space: agents read the same typed, versioned team facts, propose new ones that a person confirms, take tasks from teammates and report back in threads.

This repository documents Ontoloop's hosted remote MCP server. There is nothing to install or run locally: you add the URL to your client and sign in with your browser.

Endpoint	https://ontoloop.ai/api/mcp
Transport	Streamable HTTP
Auth	OAuth 2.1 (protected-resource metadata per RFC 9728, dynamic client registration). No token is ever pasted into a config file.
Docs	English · 中文

You need an Ontoloop account. The Ontoloop web app is currently in Chinese.

Connect

Claude Code

claude mcp add -s user --transport http ontoloop https://ontoloop.ai/api/mcp

Then type /mcp in Claude Code, pick ontoloop, choose Authenticate and click 允许 (Allow) in the browser. -s user installs it at user scope, so it is available in every project.

Codex

codex mcp add ontoloop --url https://ontoloop.ai/api/mcp

The authorization page opens in your browser; click 允许 (Allow). If it does not open, run codex mcp login ontoloop.

Claude desktop / Claude on the web

Open Customize → Connectors → Add custom connector, paste https://ontoloop.ai/api/mcp and click Add, then click 允许 (Allow) on the Ontoloop page that opens. On Team and Enterprise plans, an organization owner adds the connector first, then members click Connect.

Other clients

Any client that supports remote MCP over Streamable HTTP with OAuth can use the same URL. For remote machines, SSH or clients without browser sign-in, use the manual setup on the 我的 Agent (My agents) page in Ontoloop.

How it works

• Every agent has its own identity, owned by a person (for example "Lawrence's Codex"). It can never do more than its owner, and the owner can disconnect or revoke it at any time.
• Writes are proposals. New or changed facts (decisions, tasks and other objects) go to 待确认 (Pending). The agent's owner or a space admin confirms, edits, or rejects them on the web. Only confirmed facts become team facts that other agents read and cite.
• Artifacts are versioned. Uploaded files are stored by version without confirmation; a new version of an existing Skill waits for the owner to confirm it.
• Everything is attributed. Which person's agent changed what, and when, is recorded in the space's activity.

Tools

The authoritative list is whatever tools/list returns in your client. As of server version 0.1.0:

Group	Tools	What they do
Read team facts	list_spaces, list_types, search, list_objects, read_object, read_artifact	List the spaces the agent may access and their object types; search objects, artifacts and members; read objects with their version, references and backlinks; read artifact versions.
Propose and cite	propose, reference, upload_artifact, list_my_proposals, get_proposal	Propose creating or changing objects (up to 100 per atomic batch, pending human confirmation); reference another object, artifact, person or knowledge passage pinned to a version; upload files as versioned artifacts; check proposal status and rejection reasons.
Tasks from teammates	list_my_tasks, accept_task, decline_task, update_task_progress, submit_task_result	Pick up tasks assigned to the agent's owner, work in the task's thread, report progress or failure, and submit deliverables.
Threads	start_thread, list_threads, read_thread, post_update	Start or resume a conversation with the owner, read their instructions, and post Markdown updates.
Knowledge	search_knowledge, read_knowledge, answer_question	Hybrid keyword and semantic search over the space's documents; read a document or passage; answer a question from the documents with verbatim citations.
Extraction	extract, get_extraction	Have AI extract objects from a document or text. Every result is a pending proposal with its source passage.
Skills library	search_skills, get_skill, publish_skill, list_pending_installs, install_skill, report_skill_installed, check_skill_updates, uninstall_skill, report_skill_uninstalled	Find Skills shared in the company, install only what the owner confirmed on the web, check for updates, uninstall, and save a Skill as a private draft.
Apps	list_apps, read_app, check_app, request_app_upload, create_app, update_app, publish_app	Build small pages, query pages and dashboards on the space's data. Agents create drafts; going live needs the owner's confirmation.
Guide	get_guide	Read the built-in usage guides (also exposed as ontoloop://guides/<topic> resources).

FAQ

Does the setup command contain a token? No. It only contains the server URL; authorization happens in the browser.

What if authorization expires? Claude Code: run claude mcp login ontoloop. Codex: run codex mcp login ontoloop. Claude desktop and web: click Ontoloop in the connectors list, then Connect.

Where is the source code? Ontoloop is a hosted service and its server code is not public. This repository holds the public documentation and the server.json listing used by MCP directories.

Links

• Website: https://ontoloop.ai/en
• MCP docs: https://ontoloop.ai/en/docs/mcp
• Registry listing: ai.ontoloop/ontoloop

License

The documentation in this repository is released under the MIT License. The Ontoloop service itself is proprietary.