MCP

Sleepwalker MCP lets agents inspect Sleepwalker reports and run scoped Sleepwalker actions through the Model Context Protocol.

Use MCP when you want an AI agent to work with Sleepwalker directly from a client such as Claude, Cursor, a local MCP host, or another automation environment.

Endpoint

Hosted MCP endpoint:

https://mcp.sleepwalker.ai/mcp

Use the endpoint and connection details shown in the Sleepwalker app under MCP.

Authentication

Sleepwalker supports two MCP authentication paths:

Method	Use when
OAuth	The client supports remote MCP OAuth login.
Bearer token	The client needs an Authorization header, local proxy, script, or custom automation.

Bearer tokens start with:

sw_mcp_live_...

Tokens are user-scoped, revocable, and shown once. Store them like passwords.

See MCP setup for client examples.

What MCP Can Do

Sleepwalker MCP exposes two categories of tools.

Category	Use it for
Read tools	Inspect saved tests, reports, summaries, prompt responses, and recurring recommendations.
Action tools	Serialize pages, suggest prompts, discover trends, score content, and queue persisted runs.

OAuth connections are intended to work with the full advertised Sleepwalker MCP toolset. Bearer tokens can be narrower for custom clients and automation.

Tool Groups

Group	Tools
Saved reports	`list_sleepwalker_tests`, `get_sleepwalker_summaries`, `get_sleepwalker_test_results`, `get_sleepwalker_reports_by_url`, `get_sleepwalker_run_result`, `get_sleepwalker_prompt_response`, `get_sleepwalker_consistent_recommendations`
Page serialization	`serialize_sleepwalker_page_content`
AI Visibility	`suggest_sleepwalker_visibility_prompts`, `create_sleepwalker_visibility_run`, `list_sleepwalker_visibility_runs`, `get_sleepwalker_visibility_run_status`
Content Intelligence	`discover_sleepwalker_content_trends`, `score_sleepwalker_content`, `create_sleepwalker_content_run`, `list_sleepwalker_content_runs`, `get_sleepwalker_content_run_status`

Start with Read tools if you want an agent to inspect existing Sleepwalker data.

Start with AI Visibility runs or Content Intelligence if you want an agent to create new work.

Credits

Read tools do not consume credits.

Billable action tools return a compact billing block. Credits are shared with the API:

Action	Credits
Page content serialization	1
Prompt suggestion batch	1
Trend discovery	1
Content score with supplied trends	2
Content score without supplied trends	3
AI Visibility probe	1 per prompt/platform probe
Persisted Content Intelligence run, score depth	3
Persisted Content Intelligence run, full depth	6

See MCP billing and Credit Lifecycle for details.

Response Size

MCP clients work best with bounded responses. Sleepwalker returns compact summaries by default and provides drill-down tools for heavier data.

Use:

get_sleepwalker_prompt_response for one AI Visibility response at a time.
serialize_sleepwalker_page_content with max_chars and offset for long pages.
status tools with include_result or include_results only when you need full details.

See MCP limits for payload and polling guidance.

Common Workflow

Connect MCP
  -> list tests or create a run
  -> poll status
  -> inspect compact result
  -> fetch detailed prompt/page chunks only when needed

MCP and API use the same Sleepwalker action contract. If you build an automation in MCP first, the equivalent API route should feel familiar later.