sleepwalker

---

Sleepwalker helps teams understand how AI systems talk about a brand and what content needs to change. It combines two connected workflows:

• AI Visibility: run prompts across ChatGPT, Perplexity, Grok, and Gemini; capture full answers, citations, competitors, and mention types.
• Content Intelligence: serialize public pages, discover demand, score content depth and freshness, and return practical recommendations.

You can run the same work from the hosted app, the public API, MCP clients, or the CLI. Results stay connected, so a run started by an agent can be reviewed later in the app or queried from code.

This repository is the public developer surface for Sleepwalker: CLI package, API examples, MCP setup notes, and short guides. The hosted engine, app source, database schema, provider integrations, and billing systems are private.

Quickstart

Try the CLI without an account first. okf export is free and open source. It runs on your machine, fetches a page, and writes an agent-ready markdown bundle to a local folder. No API key, no credits:

npx -y @sleepwalkerai/cli okf export https://www.sleepwalker.ai

One command exports both concepts: the page content and the technical snapshot of what AI crawlers see. Narrow it with --content or --technical. Both concepts are described in the next section.

For everything else, install the CLI:

npm install -g @sleepwalkerai/cli
sleepwalker init

Create an API key in the Sleepwalker app, then run your first visibility check:

sleepwalker auth key set sw_api_live_...
sleepwalker doctor
sleepwalker visibility run https://yourbrand.com \
  --brand YourBrand \
  --prompt "best ai visibility platform 2026" \
  --platform perplexity,openai,grok,gemini \
  --watch

The free OKF export: content and technical

One command turns any public page into an Open Knowledge Format bundle: a folder of plain markdown files with YAML frontmatter that any AI agent can read directly. OKF is an open format published in Google Cloud's knowledge-catalog repo. The export runs entirely on your machine and is free for everyone: no account, no API key, no credits.

sleepwalker.ai-okf/
├── index.md            navigation and okf_version
├── log.md              how and when the bundle was generated
├── home.md             the content concept
└── home-technical.md   the technical snapshot

Both concepts ship by default. --content exports only the content concept, --technical exports only the snapshot.

The content concept is the page as a reader sees it: clean markdown with headings, paragraphs, lists, and links, plus the title, description, canonical URL, and timestamp in the frontmatter. Use it to hand pages to agents, prompts, and retrieval pipelines without HTML noise.

The technical snapshot is the page as an AI crawler sees it. Most AI crawlers do not run JavaScript, so the snapshot reads only the served HTML plus the response itself: the redirect chain and HTML size, curated HTTP headers, the html lang attribute, meta and title tags, canonical, icon and feed links, headings, hreflang, social tags, head scripts, every JSON-LD block, a microdata and RDFa summary, links, images, and the robots directives the page and response carry (meta robots and X-Robots-Tag). Use it to audit what ChatGPT, Claude, Perplexity, and Gemini can actually read on your pages.

One excerpt shows why this matters. A client-side rendered page often exports like this:

## Structured data

JSON-LD: none found. AI crawlers get no schema.org data from this page as served.

Everything is reported, nothing is judged: the export quotes what the page serves and leaves conclusions to you. Duplicate titles and canonicals are preserved on purpose, because they are findings. The only network request is the page fetch itself. Flags, caps, and the full section list live in docs/cli.md.

Want to see the shape of the data first? Open docs/responses.md for full example outputs from serialization, prompt suggestions, AI Visibility, Content Intelligence, and report lookup.

What you can build

• AI search (GEO) monitoring: set up prompt tracking for specific URLs across ChatGPT, Perplexity, Grok, and Gemini.
• Agent workflows: give MCP clients such as Claude access to LLM answers, cited domains, competitor performance, and content trends.
• Product integrations: use the API from internal tools, client portals, reporting pipelines, or automated QA checks.
• Content review: inspect what a page says, which trends it misses, and what to fix first.
• Free page export: turn any public page into agent-ready markdown with okf export. Open source, runs locally, no account needed.
• Free technical snapshot: every okf export also captures the page's technical layer as served, including meta tags, structured data, headers, and robots directives. --technical exports it alone.
• Custom workflows: combine Sleepwalker with other MCP servers to build your own skills and automations.

Access paths

Surface	Best for	Entry point
App	Human review, credits, keys, and full result views	app.sleepwalker.ai
API	Scripts, products, scheduled jobs, and reporting workflows	API docs
MCP	Claude and other MCP-capable agents	MCP setup
CLI	Terminal workflows and automation	CLI docs

Run Sleepwalker from anywhere

CLI

sleepwalker visibility run https://yourbrand.com \
  --brand YourBrand \
  --prompt "best ai visibility platform 2026" \
  --platform perplexity,openai,grok,gemini \
  --watch

API

curl -s https://api.sleepwalker.ai/v1/visibility/runs \
  -H "Authorization: Bearer $SLEEPWALKER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://yourbrand.com",
    "target_entity": "YourBrand",
    "prompts": ["best ai visibility platform 2026"],
    "platforms": ["perplexity","openai","grok","gemini"]
  }'

MCP

https://mcp.sleepwalker.ai/mcp

Ask an MCP-capable client: Check how YourBrand appears across AI search this week.

App

Open app.sleepwalker.ai for the visual workflow, saved results, credits, and keys.

Developer resources

Path	What it shows
docs/agents.md	MCP tool catalog and agent workflow walkthrough
docs/concepts.md	Runs, probes, serialization, scoring, credits
docs/cli.md	CLI usage, including the free local okf export
docs/cookbook.md	Runnable workflows, including CI checks
docs/responses.md	Full example outputs for the main Sleepwalker actions
mcp/server.json	Remote MCP server manifest
mcp/tools.json	Public MCP tool catalog, scopes, and credit behavior
examples/api/curl	One-call examples for public API actions
examples/api/javascript	Raw fetch plus a zero-dependency client helper
examples/api/python	urllib examples plus a small client class
examples/mcp	OAuth and bearer-token setup notes

Credits

Sleepwalker is pay as you go. Reads, lists, and status polling are normally unmetered. Actions that run work, such as visibility checks, content scoring, and serialization, use prepaid credits. okf export is the exception: it runs locally and is always free. Details live in docs/credits.md and the hosted billing docs.

Repository boundary

This repository is intentionally small.

Public here	Private in Sleepwalker
CLI, examples, setup notes, short docs	Hosted engine and app source
Public API request shapes	MCP server implementation
MCP client connection examples	Database schema and billing internals
Product-level credit behavior	Provider integrations and routing

Do not commit real keys. Use environment variables or the CLI key store:

export SLEEPWALKER_API_KEY=sw_api_live_...

Links

• Website: sleepwalker.ai
• App: app.sleepwalker.ai
• Docs: sleepwalker.ai/docs
• npm: @sleepwalkerai/cli
• Security policy: SECURITY.md
• Contributing: CONTRIBUTING.md
• License: MIT