Use case

Ask about your projects from the AI agent you already use

Most questions about a project already have an answer in Spectr: what the client agreed, which stories are waiting, how the budget looks. Spectr runs an MCP server, so you can ask them from Claude, ChatGPT, Cursor or Codex instead of opening another tab. The agent reads as you, and never sees more than your role allows.

How it works

Ask Claude. It answers from Spectr.

Claude adds Spectr's MCP address and signs in through Spectr, where the person picks which workspaces it may use. Asked "What did the client agree on Tuesday's checkout call?", it calls list_workspaces, fetch_entities, fetch_entities and answers from what it read, as that person and within their role.

The answer is in a tab you do not have open

You are halfway through an email or a pull request and need to know what was agreed on Tuesday. The answer is in a spec sheet somewhere, so you switch tools, search, and lose your train of thought. Or you guess.

What it actually looks like

A delivery lead using Claude. Three active projects.

  1. Old way

    A client asks what was agreed on the checkout call. You open Spectr, find the project, find the meeting and read the spec sheet.

  2. 9:00am

    Now: you add Spectr's MCP address to Claude once and sign in through Spectr, choosing which workspaces it can use.

  3. 11:20am

    You ask Claude what the client agreed on the checkout call. It finds the meeting and the spec sheet and answers with the story keys.

  4. 11:22am

    You ask which stories are still waiting for review. It lists three, one flagged as a possible duplicate.

  5. 4:00pm

    A developer asks Cursor for the acceptance criteria of the story they are building. Cursor reads it from Spectr, as them.

A question that took a tab switch and a search now takes one sentence in the tool you were already using.

Reads as you, never more

Every call is checked against your current role in the workspace and returns no more than the same page in Spectr shows you. You choose which workspaces a connection can reach.

Acts only with access you approve

With read and write access, an agent can take the same actions as Spectr's own buttons, as you: approve a story, refresh a budget, start a spec sheet from a brief. Actions that start AI work count toward the workspace's usage the same way the button does.

Reading is free

Reading your Spectr data over MCP costs nothing on any plan. There is no API key to create; the agent signs you in through Spectr in your browser.

Disconnect at any time

Disconnect an agent from User Settings > Connections and it is stopped on its very next request. A password reset disconnects every agent.

Works with

ClaudeChatGPTCursorCodex

Frequently asked questions

Anything else? Email hello@crowdlinker.com and a person will answer.

Which agents work with Spectr?

Any agent that supports remote MCP servers with OAuth, including Claude (claude.ai, the desktop app and Claude Code), ChatGPT, Cursor, Codex and VS Code. The setup for each is in the docs.

Can an agent see other people's projects?

Only what you can see yourself, in the workspaces you allowed when you connected it.

What does the agent's AI provider see?

What the agent reads from Spectr is sent to that agent's provider under your own agreement with them.

Also worth reading

Development with coding agents

Hand stories and tasks to a coding agent from a workstream, and track the run on the board

Meeting notes to tickets

Import a call, get a draft backlog, approve what is right

Writing user stories

Draft stories with real acceptance criteria, then approve, edit or reject each one

What teams use Spectr for →

AI ProductOps for the entire team.

Start for free at spectr.pm
  • Free plan
  • No credit card to sign up
  • 30-day free trial on Starter