How to Use Peec AI MCP with Claude Cowork

Last updated: 28 May 2026

Short answer: use Peec AI MCP with Claude Cowork for supervised reporting and team handoff, not unsupervised publishing. Cowork is a good fit for recurring visibility briefs, competitor movement summaries, and source-gap action queues that a human reviews before sending.

Peec's MCP docs say the server works with Claude surfaces and other MCP-compatible clients. Anthropic's connector docs say remote MCP connectors are available across Claude surfaces including Cowork, depending on plan, admin settings, and connector availability. Treat that as a deployment dependency: verify your workspace before promising a recurring workflow.

Best Cowork workflow

Cowork is useful when the output is a business briefing:

• Monday: run a 7-day weekly visibility pulse.
• Compare ChatGPT, Perplexity, Gemini, Claude, Copilot, Grok, Google AI Overviews, and Google AI Mode where tracked.
• Summarise competitor movers and source changes.
• Pull Peec Actions into owned, editorial, reference, and UGC buckets.
• Draft a Slack or Teams-ready report for human review.

Setup pattern

1. Add or approve the Peec connector in the Claude/Cowork environment.
2. Authorize with the right Peec account.
3. Ask Cowork to list Peec projects and confirm the selected project.
4. Save a recurring briefing prompt that uses 7 days for weekly reports.
5. Require Cowork to cite the Peec report dimensions and date range.
6. Review before sending anything to a client or team channel.

Prompt template

Run a weekly Peec AI visibility briefing for <project>.
Use the past 7 days. Break down results by engine and topic where available.
Show visibility and share of voice as percentages. Show retrieval_rate and
citation_rate as rates. Resolve all names; do not show raw IDs. Include:
1. wins, 2. losses, 3. competitor movement, 4. source changes,
5. top owned/editorial/reference/UGC actions, 6. what needs human approval.

What Cowork should not do

Cowork should not create/delete Peec prompts, publish articles, send client updates, or change website files without explicit approval. It should stage the evidence and draft the action list. Claude Code or a website agent can handle implementation after the strategy is approved.

This is the same operating discipline Schmitdy uses in an AI Search SPRINT: measurement first, source evidence second, implementation third, and desktop/mobile QA before a preview is called ready.

Source and fact ledger

• Peec AI MCP setup guide: https://docs.peec.ai/mcp/setup
• Peec AI MCP prompts: https://docs.peec.ai/mcp/prompts
• Anthropic custom connector help: https://support.anthropic.com/en/articles/11175166-getting-started-with-custom-integrations-using-remote-mcp
• Claude Cowork product page: https://www.anthropic.com/product/claude-cowork