Practical overview
When it fits
Use AEGIS when an AI agent can invoke tools and the team wants policy enforcement, human approval for selected calls, or cryptographic records of decisions before execution.
Who it is for
- Teams building AI agents that call tools and need pre-execution policy decisions.
- Operators who need human approval for selected agent actions or an audit record tied to agent identity.
- Developers with OpenAI, Anthropic, Mistral or Gemini SDK integrations who want to route calls through a policy gateway.
Documented uses
- Evaluate a tool call against policies before it executes.
- Require human review for selected calls.
- Associate calls with a stable agent identity.
- Record and verify cryptographic audit evidence for decisions.
- Route an existing supported model SDK through a gateway.
Requirements & limits
The official terms page describes a free tier and paid hosted plans. The site describes the source code as MIT licensed.
- Choose between hosted gateway, Docker Compose self-hosting and in-process Python or JavaScript/TypeScript SDK integration.
- The documented hosted SDK route is described as requiring no application-code edits; the self-hosted option brings up a gateway, cockpit, Postgres and Redis.
- Policies can validate tool-call argument shapes, while tenant DSL rules add conditions such as requiring approval for high-risk calls at specified times.
- The documentation describes signed agent identity tokens and a Merkle audit log with Ed25519-signed tree heads.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Route an existing model SDK through the hosted gateway
Configure an existing OpenAI, Anthropic, Mistral or Gemini SDK integration to use the AEGIS gateway without changing application code.
Before you start
- An existing supported model SDK integration.
- An AEGIS API key and the relevant upstream model-provider key.
Steps
Set the SDK's base URL to the documented AEGIS gateway URL for the provider.
Set AEGIS_API_KEY to the AEGIS key.
Leave the existing application code in place; the documentation says the upstream key is forwarded unchanged.
Expected outcome
The SDK is configured to send requests through the AEGIS gateway for policy evaluation before tool calls execute.
Guide limitations
- The documentation does not provide deployment-specific configuration details. It describes this as a hosted integration path.
Start the self-hosted AEGIS stack with Docker Compose
Bring up the documented AEGIS gateway and supporting services on a host or in a VPC.
Before you start
- Docker Compose.
- The AEGIS source repository.
Steps
Clone the AEGIS repository.
Change into the Aegis directory.
Run the documented Docker Compose command using docker/compose.yml.
Expected outcome
The documentation says the gateway is available on port 8080 and the cockpit on port 13003.
Guide limitations
- The guide documents stack startup, not subsequent policy configuration or an end-to-end protected call.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 7 November 2026
- Original website title
- AEGIS — a control point for AI agents' tool calls · Official homepage title · retained in its source language