Thor MCP Server for ChatGPT
By Thor
Overview
Thor connects ChatGPT to the tools your team uses, infers projects, decisions, and commitments, and enriches your agents with up-to-date context. This gives ChatGPT a single MCP connection to efficiently pull the latest data from your team and keep everyone in sync. Use Thor to: • Help ChatGPT get the freshest picture of your team’s decisions before doing its work • Create high-quality issues, check for duplicates automatically, and link to supporting work • Extract action items, decisions, and commitments from meeting transcripts Thor can connect to: • Issues in GitHub and Linear • Team chats in Slack and Microsoft Teams • Meetings in Google Meet, Zoom, and Microsoft Teams • Documents in Notion
How to connect
- ChatGPT: Open the Thor listing in the ChatGPT app directory and choose Connect. ChatGPT listing Thor exposes 17 tools here.
Tools
apply_work_item_feedback
ChatGPTcreate_issue
ChatGPTowner/repo (e.g. sindresorhus/awesome) or a github.com URL, or a Linear team key / display name. Omit repo to use the caller's default repository memory when one exists. Prefer this over ask_thor for issue creation. If the ref cannot be resolved, call search_repositories with a looser query, then retry with one of those ref values (full owner/repo when the hit is GitHub). Title and description are required. The title is used as a strong hint; the shared issue_agent still formats the final title and body (including any repository issue template). By default this checks for similar open issues in the target repository. If any are returned, show them to the user and retry with the same arguments plus skip_duplicate_check=true` only after they confirm.create_session
ChatGPTuuid4, records an initial session owned by the current principal, and returns the id as a string. Store it and pass it back as a session_id argument on later calls to persist state across a session — only an id created this way resolves. State is keyed by the authenticated principal, so the id organizes sessions within a user; on an unauthenticated connection the id is the only thing standing between callers, which is why it is unguessable.end_session
ChatGPTget_answer
ChatGPTthor_troubleshooter or ask_thor (with agent_profile) run that returned "still running", or that your tool timeout cut: the run kept going and its answer is kept for a day. With session_id, the session's latest run; with run_id, that run (from the reply that started it); with neither, your own latest run, whichever tool started it. Waits up to wait_seconds (default 100) for the answer with a progress notification every 20 s, then returns it in the same shape the starting tool would have (Markdown, then a ``json block with session_id and run_id), or "still running" again, in which case call once more. Pass wait_seconds=0` to just check. Args: session_id: The session the run was started on. run_id: The run itself, when it was started without a session. wait_seconds: How long to wait in this call for the answer, 0 or more; default 100.get_issue_context_pack
ChatGPTrepo + issue_number). A pack assembled in the last hour is served from cache. Pass refresh=true to regenerate. Optionally filter to named pack ## sections (e.g. ["Related work", "Decisions"]`).get_meeting
ChatGPTget_standup
ChatGPTget_work_item_evidence
ChatGPTlist_my_work_items
ChatGPTsearch_work_item_evidence. Work items are linked to the user via meeting transcript ownership, thread summary ownership, or authored issue/PR activity.search_meetings
ChatGPTsearch_repositories
ChatGPTref straight back to create_issue(repo=...). Prefer the full GitHub owner/repo` ref. Do not invent UUIDs.search_work_item_evidence
ChatGPTscope=viewer to search only YOUR linked items; scope=organization for org-wide recent activity. scope=profile requires an active agent profile and errors on this MCP tool. Use for blockers / ready-to-close / headline match. Do not use this to list your plate — that is list_my_work_items. Prefer this for exploratory work analysis. Follow up with get_work_item_evidence for details. Do not use get_standup for general "what did people work on" questions. Usage: - "Find the outreach videos item" -> query="outreach videos" (no finding_kinds) - "What's blocked?" -> finding_kinds=["active_blocker"] - "What looks ready to close?" -> finding_kinds=["provider_open_but_work_done"] - "Blockers on my plate" -> scope="viewer", finding_kinds=["active_blocker"] Do not pass finding_kinds on a named lookup. Kinds are a triage filter after you have a set, not part of finding a title. Viewer search without query/queries/finding_kinds is a worse list_my_work_items. status is the provider-backed read model. When a row has status_conflict, a human override disagrees — report the read-model status and explain the contradiction. Do not tell the user the item is the human status.send_standup_summary
ChatGPTthor_data_status
ChatGPTthor_onboarding
ChatGPTCapabilities
Links
Similar Apps
Listed on MCP App Store
Add this badge to your site or README. It links to this page.
HTML
<a href="https://mcp-app-store.keenethics-labs.com/app/thor"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>
Markdown
[](https://mcp-app-store.keenethics-labs.com/app/thor)






