The AI-native knowledge base for GitHub Copilot

The GitHub Copilot MCP server for long-term memory. Copilot is great inside the file. Hjarni is where the project context lives.

Get started free See how it works

Free to start. No credit card required.

What this unlocks

Workflows that actually use your context

Ask Copilot Chat the questions it cannot answer from code

"Why did we ditch the background job approach for billing?" Copilot searches Hjarni for the ADR, reads it, and answers from your actual reasoning instead of inventing one.

Give Copilot long-term memory before agent mode writes

Before agent mode writes the change, it pulls your Hjarni conventions notes: error handling style, test naming, what the team rejected last time. The diff lands closer to mergeable on the first try.

Skip the architecture preamble in every chat

If you keep pasting the same paragraph about how the services fit together, save it as a Hjarni note once. Copilot reads it from MCP on every new conversation.

Write the runbook while the incident is fresh

End an incident by asking Copilot to draft a Hjarni note: what failed, what you tried, the actual fix, and what to check next time. The next on-caller's Copilot reads it back.

One knowledge base across Copilot, Claude, and Cursor

Hjarni is account-scoped, not editor-scoped. The notes you write while pair-programming with Copilot are the same notes Claude reads in the browser and Cursor reads in a side project.

Setup

Connect GitHub Copilot in about two minutes

1
Sign up for Hjarni and write a few notes describing how the codebase is meant to fit together.
2
In VS Code, create .vscode/mcp.json in the repo, or open the command palette and run "MCP: Open User Configuration" for a user-wide config.
3
Paste the JSON above. The root key is "servers" (not "mcpServers"), which is the VS Code shape.
4
Open Copilot Chat in agent mode. VS Code opens a browser tab for the Hjarni OAuth flow. Sign in once.
5
In the agent chat, confirm the Hjarni tools appear in the tool list, then ask Copilot to search Hjarni for one of your notes.

.vscode/mcp.json (workspace) or user mcp.json via MCP: Open User Configuration

{
  "servers": {
    "hjarni": {
      "type": "http",
      "url": "https://hjarni.com/mcp"
    }
  }
}

The Hjarni MCP endpoint is https://hjarni.com/mcp. The connection uses OAuth, so you sign in with your Hjarni account when prompted. No API key files to manage.

Why a GitHub Copilot MCP knowledge base sits alongside repo indexing

Copilot is excellent at reading the open file and the surrounding repo. It is not designed to read the design doc in someone's Notion, the decision buried in a Linear comment, or the runbook a teammate keeps in their Apple Notes. Hjarni is one place all of that can land, with MCP search built into Copilot Chat through the same mcp.json file VS Code already understands.

Agent mode is the highest-leverage moment to inject context. If Copilot starts a multi-step task with the wrong assumptions, every step is wrong. Folder-level AI instructions in Hjarni let you steer Copilot per project: 'in the Payments folder, never propose schema changes without an ADR'. The instructions live with the notes, not in the editor, so they survive every Copilot update.

The same Hjarni knowledge base is reachable from Copilot in VS Code, Claude in the browser, ChatGPT on your phone, and Cursor on a side project. The notes you write today work in the editor you switch to next year. One source of truth, every AI surface.

Common questions

Questions before you connect GitHub Copilot

Where does the Copilot MCP config file live?

VS Code reads two locations. Per-project, drop the config in .vscode/mcp.json inside the repo. User-wide, open the command palette and run "MCP: Open User Configuration", and VS Code opens the user-level mcp.json for you. Note the root key is "servers", not "mcpServers" like Cursor and Claude Desktop.

Which Copilot plan do I need for MCP?

MCP servers are surfaced through Copilot Chat agent mode in VS Code, which is included on Copilot Pro and above. Business and Enterprise admins can additionally restrict which MCP servers their members may add, so on those plans you may need a policy toggle from your admin before Hjarni shows up. Coverage on Copilot Free moves around, so check GitHub's current Copilot docs for the up-to-date matrix before you build a workflow around it.

Copilot already indexes my repo. Why add Hjarni?

Copilot sees code. Hjarni stores everything that is not in the repo: architecture decisions, runbooks, customer feedback, past incidents, and why the auth layer looks the way it does. Copilot Chat reads those through MCP when it needs context the codebase cannot give it.

Does Hjarni conflict with the official GitHub MCP server?

No. They serve different jobs. The GitHub MCP server gives Copilot access to issues, PRs, and repo metadata. Hjarni gives it access to your knowledge base. Both can be listed in the same mcp.json and Copilot will pick the right tool per question.

Does this work in Visual Studio or JetBrains, not just VS Code?

Copilot's MCP support is most complete in VS Code today. Visual Studio added MCP support more recently and JetBrains is catching up. Hjarni's endpoint is the same everywhere (https://hjarni.com/mcp with OAuth), so any Copilot host that supports remote MCP servers works.

Give GitHub Copilot a memory

Write notes once. GitHub Copilot reads them across every conversation.

Get started free See pricing

More integrations

Hjarni for Claude Hjarni for ChatGPT Hjarni for Cursor Hjarni for Claude Code Hjarni for Windsurf Hjarni for Cline

Write once. You both remember.

Free to start. No credit card required.

Give your AI a memory

Works with Claude and ChatGPT today. Gemini coming soon.