# orsel > A web action layer for AI agents. Find a page or an action on any website, read it, run it and get a typed result, watch it for changes, explain why it ran, and require a person to approve anything risky. Seven operations, the same in REST, the TypeScript and Python SDKs, the CLI and the MCP server: find, read, act, watch, explain, approve, revoke. Every run lands in a hash-chained ledger you can verify yourself. ## Start here - [Overview](https://orsel.egeuysal.com/index.md): what orsel is and its status - [Introduction](https://orsel.egeuysal.com/docs/introduction.md): What this is and how the pieces fit. - [Quickstart](https://orsel.egeuysal.com/docs/quickstart.md): Find, run and verify an action in a few minutes. - [Use a website as an API](https://orsel.egeuysal.com/docs/website-as-api.md): Turn a site into typed calls in five minutes, from the CLI, TypeScript or Python. - [Concepts](https://orsel.egeuysal.com/docs/concepts.md): Operations, risk classes, tiers, recipes, and the trust plane. - [One contract, thin layers](https://orsel.egeuysal.com/docs/adapters.md): REST, the SDKs, the CLI, MCP and agent skills are thin layers over the same operations. - [Typescript sdk](https://orsel.egeuysal.com/docs/sdk-typescript.md): The orsel package. - [Python sdk](https://orsel.egeuysal.com/docs/sdk-python.md): The orsel package for Python. - [Cli](https://orsel.egeuysal.com/docs/cli.md): Every command, the exit codes, and where config lives. - [Mcp server](https://orsel.egeuysal.com/docs/mcp.md): Find, read, act, explain and approve as MCP tools. - [Trust and proof](https://orsel.egeuysal.com/docs/trust-and-proof.md): Policy, approvals, and the ledger. - [Security model](https://orsel.egeuysal.com/docs/security-model.md): What is protected, and the known gaps. - [Site owners](https://orsel.egeuysal.com/docs/site-owners.md): Claim a domain, set policy, set prices. - [Spec](https://orsel.egeuysal.com/docs/spec.md): The manifest and recipe format. - [Webhooks](https://orsel.egeuysal.com/docs/webhooks.md): Events and how to verify them. - [Errors and limits](https://orsel.egeuysal.com/docs/errors-and-limits.md): The error envelope, codes, and rate limits. - [Migrate from another tool](https://orsel.egeuysal.com/docs/migrate.md): Move from a browser agent or a search api. ## API and tools - [OpenAPI 3.1](https://orsel.egeuysal.com/v1/openapi.json): the contract, generated from the same schemas the server validates with - [MCP server](https://orsel.egeuysal.com/mcp): Streamable HTTP, send your key as a Bearer token; tools find, read, act, approve, explain - [agents.json](https://orsel.egeuysal.com/agents.json): machine-readable description of everything on this list - [Action Manifest JSON Schema](https://orsel.egeuysal.com/spec/manifest.schema.json): what a site publishes - [Recipe JSON Schema](https://orsel.egeuysal.com/spec/recipe.schema.json): the signed unit that runs an action - [Spec](https://orsel.egeuysal.com/spec.md): manifest and recipe, with a reference validator ## Set up for an agent - [Setup prompt](https://orsel.egeuysal.com/agents/setup-prompt.md): paste into an agent to have it install and wire up orsel in a project - [orsel skill](https://orsel.egeuysal.com/agents/skill.md): when and how an agent should find, read, act, explore, approve and explain (SKILL.md); more skills are in the skills index - [Skills index](https://orsel.egeuysal.com/.well-known/agent-skills/index.json): agent skills discovery ## Registry - [Registry](https://orsel.egeuysal.com/registry.md): the sites orsel can act on, with risk class and measured reliability - [registry.json](https://orsel.egeuysal.com/registry.json): the same list as JSON; a site is at /registry/.json - [ActionBench](https://orsel.egeuysal.com/benchmark.md): reliability per site and action, read and write separately, failures published - [ActionBench methodology](https://orsel.egeuysal.com/benchmark/methodology.md): how runs are counted ### Sites - [jsonplaceholder.typicode.com](https://orsel.egeuysal.com/sites/jsonplaceholder.typicode.com.md): 2 actions - [egeuysal.com](https://orsel.egeuysal.com/sites/egeuysal.com.md): 7 actions ## Identity - [Bots](https://orsel.egeuysal.com/bots.md): how orsel identifies itself to sites, what it honors, and how to opt out - [Key directory](https://orsel.egeuysal.com/.well-known/http-message-signatures-directory): public keys that sign our requests (Web Bot Auth) ## Legal - [abuse and takedown](https://orsel.egeuysal.com/abuse.md): how to report abuse, ask for removal, or opt out. - [acceptable use](https://orsel.egeuysal.com/acceptable-use.md): what you may not do with the service. - [cookies](https://orsel.egeuysal.com/cookies.md): which cookies the site and the dashboard use. - [data processing addendum](https://orsel.egeuysal.com/dpa.md): the terms for processing personal data on a customer's behalf. - [privacy](https://orsel.egeuysal.com/privacy.md): what data the service stores, why, for how long, and who else touches it. - [security](https://orsel.egeuysal.com/security.md): how the service is secured, and how to report a problem. - [site owner terms](https://orsel.egeuysal.com/site-owner-terms.md): the terms for claiming a domain and setting policy for agents. - [subprocessors](https://orsel.egeuysal.com/subprocessors.md): the services that process data for us. - [terms](https://orsel.egeuysal.com/terms.md): the terms for using the service. ## Optional - [Everything in one file](https://orsel.egeuysal.com/llms-full.txt) - [Changelog](https://orsel.egeuysal.com/changelog.md) - [Sitemap](https://orsel.egeuysal.com/sitemap.xml) - [RSS](https://orsel.egeuysal.com/rss.xml)