# Azta > Azta answers legislation questions (tax, accounting, legal, employment and company law) with every answer linked to the provisions it cites. Use it from ChatGPT, Claude and coding agents through the MCP server at https://mcp.azta.ai/mcp, or from your product through the API at https://api.azta.ai/v1. ## Pages - [Introduction](https://docs.azta.ai/index.md): Azta answers legislation questions (tax, accounting, legal, employment and company law), and every answer links to the provisions it cites. - [Build with a coding agent](https://docs.azta.ai/agents.md): Two steps to let Claude Code, Cursor, Codex or GitHub Copilot build an Azta API integration for you. - [Claude Code](https://docs.azta.ai/agents/claude-code.md): Install the Azta API skill in Claude Code and paste one prompt to build your integration. - [Codex](https://docs.azta.ai/agents/codex.md): Install the Azta API skill in Codex and paste one prompt to build your integration. - [GitHub Copilot](https://docs.azta.ai/agents/copilot.md): Install the Azta API skill for GitHub Copilot, then use agent mode in VS Code or assign an issue to the Copilot coding agent. - [Cursor](https://docs.azta.ai/agents/cursor.md): Install the Azta API skill in Cursor and open one prompt in the agent chat. - [API reference](https://docs.azta.ai/api-reference.md): The Azta API, generated from the OpenAPI 3.1 document core publishes. - [Cancel a response](https://docs.azta.ai/api-reference/cancelResponse.md): Stop a response. A dropped stream cancels nothing; this call does. - [Create a response](https://docs.azta.ai/api-reference/createResponse.md): Ask a question and get an answer from the law, with its sources. - [Fetch provisions](https://docs.azta.ai/api-reference/fetchProvisions.md): Read provisions by `unit_id`, from a source record or a search hit: at most 20 per call. A `unit_id` names a provision, not a version, so it keeps working after the law is amended. - [Retrieve a response](https://docs.azta.ai/api-reference/getResponse.md): A response this key created, whatever its state. A failure is in the object (`status`, `error`), never in the HTTP status. - [response.cancelled](https://docs.azta.ai/api-reference/response-cancelled.md): A `background: true` response was cancelled. The event names the response and carries no answer: fetch it with `getResponse`. - [response.completed](https://docs.azta.ai/api-reference/response-completed.md): A `background: true` response completed with an answer. The event names the response and carries no answer: fetch it with `getResponse`. - [response.failed](https://docs.azta.ai/api-reference/response-failed.md): A `background: true` response failed or was declined. The event names the response and carries no answer: fetch it with `getResponse`. - [Search provisions](https://docs.azta.ai/api-reference/searchProvisions.md): Search the current provisions of the law in force, with no model call. - [Deprecation policy](https://docs.azta.ai/deprecation-policy.md): What stays stable in the Azta API and MCP server, and how much notice breaking changes get. - [Errors and limits](https://docs.azta.ai/guides/errors.md): Every HTTP error code, the failed and cancelled response states, headers, rate limits and the retry policy. - [Keys and sandbox](https://docs.azta.ai/guides/keys-and-sandbox.md): API key formats, how a key is shown, and the free sandbox with its sample sources. - [Lookups](https://docs.azta.ai/guides/lookups.md): Search provisions and fetch their text directly, without running an answer. - [Use the OpenAI SDK](https://docs.azta.ai/guides/openai-sdk.md): Call the Azta API from the official OpenAI Python and Node SDKs, and keep the SDK's automatic retries safe with a per-call Idempotency-Key. - [Pricing](https://docs.azta.ai/guides/pricing.md): What the API charges per answer and lookup, which outcomes are free, and how top-ups work. - [Responses](https://docs.azta.ai/guides/responses.md): Create a response, stream it or poll it, cancel it, ask follow-up questions and retry safely. - [Sources](https://docs.azta.ai/guides/sources.md): The source record every answer carries, how its url is chosen, and how to render citations from annotation offsets. - [Webhooks](https://docs.azta.ai/guides/webhooks.md): Get a signed webhook when a background response ends, verify it with Standard Webhooks, and fetch the answer. - [Use Azta in ChatGPT, Claude and coding agents](https://docs.azta.ai/mcp.md): Connect the Azta MCP server to ChatGPT, Claude, Cursor, VS Code, Claude Code, Codex and Gemini CLI. - [Quickstart](https://docs.azta.ai/quickstart.md): Ask Azta one question over the API and read the answer with its sources, in curl, Python or TypeScript. - [Support](https://docs.azta.ai/support.md): How to reach Azta support. - [Trust](https://docs.azta.ai/trust.md): Where Azta API data is processed and stored, what is never used for training, how long it is kept, and which documents are binding. - [Data use](https://docs.azta.ai/trust/data-use.md): What Azta and its processors do with your questions, documents and answers. - [Data processing agreement](https://docs.azta.ai/trust/dpa.md): Where to find the data processing agreement that applies to your use of the Azta API. - [Hosting and processors](https://docs.azta.ai/trust/hosting.md): Where Azta stores data, where its models run and where web searches go. - [Retention](https://docs.azta.ai/trust/retention.md): How long Azta keeps API key runs, webhook deliveries and request logs. ## Other pages - [Changelog](https://docs.azta.ai/changelog.md) - [Status](https://docs.azta.ai/status.md) ## Agent skill - [Azta API skill](https://docs.azta.ai/skills/azta-api.md): install page for the `azta-api` agent skill - [Skill archive](https://docs.azta.ai/skills/azta-api.tar.gz): the skill as a gzipped tar - [Discovery index](https://docs.azta.ai/.well-known/agent-skills/index.json): Agent Skills discovery 0.2.0, with the archive digest