Skip to content
Spekir
LiveMCP connector

Atlas in your developer's IDE

Engineers query application ownership, ADRs, and technology decisions inline in Cursor — without leaving the editor or opening a browser tab. Atlas answers from the real EA data, not from memory.

Get started →

What engineers use it for

Ownership lookup

@atlas get_app_context billing-service

Type the app name while reviewing code, and Atlas returns the owner team, lifecycle status, business criticality, and the top 3 business capabilities it supports — all inline in the editor. No browser tab required.

ADR lookup before you build

@atlas find_decisions authentication

Before writing an auth implementation, ask Atlas for relevant ADRs. You get the decision text, status, and a direct link to the full record in Atlas. Catch existing decisions before reinventing solutions.

Portfolio check before adding a service

@atlas list_applications {"lifecycle": "active"}

Before proposing a new microservice, pull the active application portfolio. Check whether a service already exists, what its TIME classification is, and who owns it. Architecture decisions grounded in the real portfolio.

Set up in 3 steps

Requires Cursor 0.40+ and an Atlas workspace.

1

Create an MCP API key in Atlas

Go to atlas.spekir.com → Settings → MCP API Keys. Click Create API key, give it a name, and select read:* scopes. Copy the key — it is shown once in plaintext.

2

Add Atlas to ~/.cursor/mcp.json

Create or edit ~/.cursor/mcp.json and add the Atlas server block with your API key as the Authorization header. Restart Cursor after saving.

{
  "mcpServers": {
    "atlas": {
      "url": "https://atlas.spekir.com/api/mcp/v1",
      "headers": {
        "Authorization": "Bearer mcp_YOUR_KEY_HERE"
      }
    }
  }
}
3

Query Atlas inline

In Cursor Chat, type @atlas followed by any tool name. Start with @atlas list_applications to confirm the connection works.

Full setup guide with troubleshooting →

Your data stays yours

Atlas MCP API keys are workspace-scoped. A key can only return data from the specific Atlas workspace it was created in — no cross-workspace leakage is possible at the query layer. Keys can be revoked instantly from Settings → MCP API Keys.

Read-only by design

All v1 tools are read-only. No data is written to your Atlas workspace from Cursor queries. Proposed architecture changes go through the full Atlas change-proposal review flow before anything is committed.

Frequently asked

Which Cursor version is required?

Cursor 0.40 or later. MCP server support was introduced in 0.40. Update via Cursor → Check for Updates if you're on an older version.

Is my architecture data safe?

Atlas MCP API keys are workspace-scoped — they only return data from your specific Atlas workspace. Keys can be revoked instantly from Settings → MCP API Keys. No architecture data is stored by Cursor.

Do all team members need their own API key?

Yes. Each developer should create their own key so you can revoke per-person access independently. Keys are rate-limited per key (60 req/min default), so sharing a key will exhaust the limit faster.

Can Atlas write changes back from Cursor?

No. All v1 tools are read-only. Write tools (propose_changes, get_proposal_status) are planned for v2 — any changes proposed from the IDE will go through Atlas's existing change-proposal review flow before anything is committed.

What if my team uses Windsurf or another AI editor?

Atlas uses standard MCP Streamable HTTP transport (JSON-RPC 2.0 over POST). Any MCP-compatible client can connect using the same configuration. Contact hello@spekir.com if you need help with a specific editor.

Ready to connect Atlas to Cursor?

Get started →Talk to us →