# EgisAI

> Runtime governance for AI agents — install the egisai SDK and your OpenAI, Anthropic, and Google calls are gated by your organization's policies automatically.

## Docs

- [egisai.init()](https://docs.egisai.co/api-reference/init.md): Activate the SDK for the current process. Idempotent.
- [API reference](https://docs.egisai.co/api-reference/overview.md): Public symbols exposed by the egisai package.
- [egisai.set_context()](https://docs.egisai.co/api-reference/set-context.md): Attach request-scoped metadata to every governed call inside the current task or thread.
- [egisai.shutdown()](https://docs.egisai.co/api-reference/shutdown.md): Stop background workers and flush remaining audit events.
- [Types](https://docs.egisai.co/api-reference/types.md): Public dataclasses and enums returned by the policy primitives.
- [Agents](https://docs.egisai.co/concepts/agents.md): How the SDK identifies which agent made each call, and how to override that identity for sub-agents.
- [How it works](https://docs.egisai.co/concepts/how-it-works.md): How the egisai SDK slots into your model call path — evaluation, outcomes, and telemetry.
- [Policies](https://docs.egisai.co/concepts/policies.md): Categories of rules supported by the SDK, and how the SDK consumes the policies your operators configure.
- [Verdicts](https://docs.egisai.co/concepts/verdicts.md): Every governed call resolves to one of three verdicts: allow, sanitize, or block.
- [Blocking behavior](https://docs.egisai.co/guides/blocking-behavior.md): Choose between raise and stub modes, and handle blocked calls cleanly.
- [Configuration](https://docs.egisai.co/guides/configuration.md): Every option egisai.init() accepts, every environment variable it honours, and how to combine them.
- [Multi-agent context](https://docs.egisai.co/guides/multi-agent-context.md): Run several distinct agents from the same process and attribute every call correctly.
- [Troubleshooting](https://docs.egisai.co/guides/troubleshooting.md): Symptoms, what they mean, and how to fix them.
- [Installation](https://docs.egisai.co/installation.md): Install the egisai SDK with the right extras for your stack and verify the install.
- [Anthropic](https://docs.egisai.co/integrations/anthropic.md): Govern Anthropic Messages API calls — including streaming — without code changes.
- [Google Gemini](https://docs.egisai.co/integrations/google.md): Govern Google Generative AI calls — including streaming — without code changes.
- [httpx and requests](https://docs.egisai.co/integrations/http-clients.md): Optional HTTP-level capture for libraries that don't go through one of the official provider SDKs.
- [OpenAI](https://docs.egisai.co/integrations/openai.md): Govern OpenAI Chat Completions, Responses API, and streaming calls without code changes.
- [EgisAI](https://docs.egisai.co/introduction.md): Runtime governance for AI agents — install the SDK, call egisai.init(), and your existing OpenAI, Anthropic, Google Gemini, or HTTP-based calls are governed automatically.
- [Quickstart](https://docs.egisai.co/quickstart.md): Install egisai, initialize once, and watch your first governed call land on the dashboard.
- [Releases](https://docs.egisai.co/releases.md): Versioning policy, the changelog, and how to verify a downloaded wheel.
- [Privacy and security](https://docs.egisai.co/security.md): How egisai handles sensitive data, what to keep out of source control, and where to report vulnerabilities.
- [Support](https://docs.egisai.co/support.md): Where to ask questions, file bugs, and report security issues.

## OpenAPI Specs

- [openapi](https://docs.egisai.co/api-reference/openapi.json)