The Fleece AI Brain GitHub connector syncs the issues and pull requests across every repository you are involved in into a local-first knowledge base — plain Markdown and SQLite on your own machine. Each item becomes a note titled by its repo and number, and any MCP-compatible app such as Claude Desktop or Cursor can answer from your real GitHub history, with the source one click away.
What syncs from GitHub
Everything arrives as knowledge nodes — ordinary files you can open, link and search, not rows locked inside someone else's cloud.
Issues
Every issue you created, were assigned, were mentioned in or are subscribed to, across all your repositories, pulled in through GitHub's issues feed.
Pull requests
Pull requests come through the same feed and are tagged type/pr, so the discussion behind a merge is remembered alongside the issues.
Bodies as notes
Each item becomes a note titled "owner/repo #123: title", with the description and the canonical GitHub URL in the body.
Repo and type tags
Notes are tagged github, type/issue or type/pr, and repo/owner-repo, so you can filter your brain by project or item type.
Incremental sync
Runs resume from a stored cursor (the newest updated_at seen); the first sync backfills the last 90 days rather than your whole history.
Update-in-place & dedup
A seen-ledger keyed on the issue id means an edited item refreshes its note in place and a re-sync never creates duplicates.
Connect GitHub in four steps
No CI job, no webhook plumbing — one sign-in from the desktop app.
- 01
Install Fleece AI Brain
Download the desktop app for macOS, Windows or Linux and open your vault — a folder of plain Markdown files on your machine.
- 02
Open Connectors and pick GitHub
In the app, open the connectors panel, choose GitHub, and sign in with your account. OAuth is brokered securely; your GitHub content never passes through our servers.
- 03
Let it scope to your work
The connector reads the issues and pull requests you are involved in — assigned, created, mentioned or subscribed. Nothing outside your own activity is pulled.
- 04
Let it sync — then ask
Sync runs in the background and stays up to date. Point Claude Desktop, Cursor or any MCP app at your Brain and ask why a change was made months ago.
The code ships; the reasoning gets buried.
A pull request captures why a change was made — the trade-off weighed, the edge case caught, the approach rejected. An issue captures what a bug actually was and how it was decided. Then both scroll out of view, and the next engineer, or the next coding agent, starts from the diff alone.
The GitHub connector ends the re-archaeology. Issues and pull requests become permanent notes in one living knowledge graph, linked to the repositories, people and decisions they mention — so the reasoning from a PR merged in March is still one question away in December.
Your issues go from GitHub to your disk. Nowhere else.
Authentication is a standard OAuth sign-in, but the actual sync calls the GitHub REST API directly from the desktop app on your machine. Raw issues and pull requests land straight in your vault as Markdown and SQLite — they don't transit Fleece's servers.
That is the whole trust model: we can't read your GitHub history, period. The files are yours, they open in Obsidian or any editor, and if you ever leave, you take everything with you.
- +Direct GitHub-to-device sync — raw payloads never touch our servers
- +Vault = plain .md files + SQLite, 100% Obsidian-compatible
- +Works offline once synced; cloud sync is optional and end-to-end encrypted
Give your coding agents the context behind the code.
One copy-paste connects the Brain to Claude Desktop, Cursor, Cline, Zed or agents you build. From then on, "why does this module handle retries this way?" is answered from the actual issue or PR thread — with the source note one click away.
A new engineer and a coding agent read the same graph, so neither starts blind. On Teams, the organization map also shows which agents used that knowledge and what they cost.
GitHub connector — FAQ
Does my GitHub data pass through Fleece's servers?+
No. Sign-in uses a brokered OAuth flow, but the sync itself calls the GitHub REST API directly from the desktop app on your machine. Raw issues and pull requests land straight in your local vault; we never see or store your GitHub content.
What exactly does the connector sync, and can I limit it?+
It syncs the issues and pull requests you are involved in — created, assigned, mentioned or subscribed to — across your repositories. It is already scoped to your own activity through GitHub's issues feed, so it never pulls repositories you have no part in.
Does it write anything back to GitHub?+
No. The connector is read-only: it pulls issues and pull requests into your knowledge base and never opens, comments on, edits or closes anything in your repositories.
How often does it sync, and does it refetch everything?+
Sync runs in the background and resumes from a stored cursor, so each run only fetches recently updated items; the first sync backfills the last 90 days. An edited issue refreshes its note in place, and a re-sync never creates duplicates.
Which plan do I need to connect GitHub?+
The local brain is free, and connectors are the paid lever: Solo (€12/month) includes up to 3 connected tools, Pro (€24/month) makes them unlimited with auto-sync. Every plan starts with a 14-day trial, no card required.
How do I ask Claude or Cursor about my GitHub history?+
Connect the Brain to Claude Desktop, Cursor or any MCP-compatible app with a single copy-paste. Your AI then answers from the synced issues and pull requests, citing the exact note it drew from.
Can I open the synced issues outside the app?+
Yes. Every issue and pull request is an ordinary Markdown file in your vault, fully Obsidian-compatible — open, edit or grep them with any tool, and take them with you if you ever leave.
What happens if I disconnect GitHub?+
Sync stops, and everything already in your vault stays yours — the notes are local files, not a mirror that vanishes with the connection.
Stop shipping code whose reasoning is lost.
Connect GitHub in one sign-in and give every engineer and coding agent the context behind the code.