How is workspace MCP different from customer-domain MCP?

Workspace MCP is an authenticated service for querying authorized KnitKnot data and updating supported playbook states. Customer-domain MCP is a rollout-gated, read-only interface over published managed artifacts.

Does MCP expose every KnitKnot operation?

No. Customer-visible tool discovery uses an explicit allowlist. Dense configuration, benchmark execution, and report publishing remain console workflows.

Do supported MCP updates appear in the console?

Yes. The assistant and console use the same authorized workspace records, so a supported playbook status change appears in both places.

Does Google Analytics capture every AI-influenced visit?

No. It records observed referral sessions when a recognized source is present. KnitKnot keeps that directional traffic signal separate from benchmark scoring.

# KnitKnot workspace MCP and integrations

An MCP-capable assistant can read an authorized KnitKnot workspace directly: issues, their evidence, playbooks, sources, and demand topics. Notion, GitBook, and Google Analytics connect separately, each for one job.

## What an assistant can do with a connected workspace

KnitKnot exposes a curated set of tools over the same authorized workspace data used by the console. Supported tools cover competitive position, issues and their evidence, playbooks, sources, demand topics, questions, coverage, and workspace context. Supported playbook status changes appear in the console.

Interactive MCP clients can use OAuth 2.1. Scripts and other non-interactive clients can use a KnitKnot API key. Every request is scoped to the caller's authorized workspace, and customer-visible tools are controlled by an explicit allowlist. Benchmark runs, dense configuration, and report publishing remain console workflows.

## Know which MCP surface you are connecting to

This authenticated workspace service is separate from the rollout-gated read-only MCP interface on managed customer domains. The customer-domain interface reads published page, search, and approved-fact artifacts.

## Which integration covers which job

- Notion coordinates the playbook backlog in a connected database.

- GitBook stores the configured documentation destination.

- Google Analytics adds observed AI-referred sessions and landing pages beside benchmark trends. Referral data is directional and remains separate from benchmark scoring.

## Related pages

MCP & Integrations

Query your KnitKnot workspace over MCP, and connect the tools around it.

Product evidence plate

Workspace MCP Authenticated · workspace scoped

Request

Which issue reaches the most buyer answers, and what work is linked to it?

list_issues

Returns the ranked issue index for the authorized workspace.

get_issue

Opens the selected issue, its evidence, and linked playbooks.

AI repeats an outdated deployment limitation

14 affected answers · 2 linked playbooks · evidence available

Endpoint

https://mcp.knitknot.ai/mcp/

Authentication

An interactive client approves through OAuth 2.1. A script or other non-interactive client uses a KnitKnot API key.

Scope

Every request resolves against the caller's authorized workspace.

Tool discovery

Customer-visible tools are controlled by an explicit allowlist. Benchmark runs, dense configuration, and report publishing remain console workflows.

Read the MCP setup guide

Surface	Reads	Access	Primary use
Workspace MCP	Authorized workspace evidence and supported work records	OAuth 2.1 or API key	Research and supported playbook coordination
Customer-domain MCP	Published pages, search artifacts, and approved facts	Read-only and rollout-gated	Agent access to the published corpus

Notion: Coordinate the playbook backlog; Use a connected database for team review while supported playbook status changes stay tied to the KnitKnot record.
GitBook: Set the documentation destination; Validate the connection and select the workspace destination used by the documentation workflow.
Google Analytics: Add observed AI referral data; Show recognized AI-referred sessions and landing pages beside the benchmark without adding them to the score.