How it worksWho it's forBlog
Nazr vs LinearThe plan above the tracker, kept trueNazr vs JiraA living, reviewed plan vs a visibility layer
Open app
← Back to Nazr

Developers

Connect your AI to Nazr

One MCP endpoint. Claude, ChatGPT, Claude Code, or any MCP client.

Nazr exposes a Model Context Protocol endpoint at https://app.nazr.com/mcp. A connected agent pulls real tasks with full context, reports what it shipped, and raises signals when the world disagrees with the plan. It can never change the plan directly — every proposed change lands in your review queue, and only humans accept.

Claude (claude.ai)

  1. Open Settings → Connectors → Add custom connector.
  2. Enter https://app.nazr.com/mcp as the URL — no keys needed.
  3. Claude sends you to Nazr to sign in and approve the connection (OAuth). Approve, and you're connected.

Then try: "What should we work on next?" — or after a good strategy conversation, "Send this conversation's decisions to my roadmap as signals." Claude distills the decisions; each arrives in your review queue with provenance.

Claude Code

claude mcp add --transport http nazr https://app.nazr.com/mcp

Claude Code walks the same OAuth consent in your browser. For headless or CI use, create an API key in Nazr under Integrations → Your agent and pass it instead:

claude mcp add --transport http nazr https://app.nazr.com/mcp --header "Authorization: Bearer nazr_…"

ChatGPT

  1. Enable Developer mode (Settings → Apps & Connectors → Advanced).
  2. Add a connector with MCP server URL https://app.nazr.com/mcp and OAuth authentication.
  3. Sign in to Nazr and approve when prompted.

Any other MCP client

The endpoint speaks Streamable HTTP with OAuth 2.1 (dynamic client registration, PKCE) or Authorization: Bearer API keys. Discovery metadata lives at https://app.nazr.com/.well-known/oauth-protected-resource/mcp.

What your agent can do

ToolAccessWhat it does
get_planReadThe full plan tree with statuses and reconcile state
next_taskReadThe highest-leverage startable task, as a complete brief
get_taskReadOne task's brief by id
start_taskWriteMark a task in progress
complete_taskWriteReport completion — summary, steps, artifact links
block_taskWriteReport a blocker with the reason
raise_signalWriteSend an observation or decision to the review queue
get_ground_truthReadThe org's ground-truth fact ledger
record_ground_truthWriteRecord or supersede a durable fact

Trust & revocation

  • Humans approve every plan change. Signals and completions from agents become reviewable proposals — never silent edits.
  • Org-scoped. A connection acts inside one organization, as the user who approved it.
  • Revocable. Disconnect from your MCP client, revoke API keys in Nazr under Integrations, and your data handling is covered by our Privacy Policy.
How it worksWho it's forBlogNazr vs LinearNazr vs JiraPrivacyTermsOpen app
Nazr AI LtdCompany no. SC892023hello@nazr.com