About AgentRail
AgentRail is an open-source flight recorder for AI agents. It records traces, tool actions, cost metadata, actor attribution, and evidence payload status so developers can debug and audit autonomous AI workflows.
Startup status
AgentRail is an early open-source startup project built from Indonesia. The current milestone includes a TypeScript SDK, ingestion pipeline, worker-side pricing, local Docker setup, forensic dashboard, guided demo, and read-only MCP server.
The product is seeking founding testers before hosted multi-tenant work. The immediate goal is to prove that AI agent traces can be captured, priced, and investigated without putting sensitive evidence directly into chat clients.
30/60/90 day roadmap
- 30 days
- Finish package publish readiness, collect feedback from three founding testers, and cut the first public SDK/MCP release.
- 60 days
- Document an AWS reference deployment using API Gateway, Lambda, SQS, RDS/PostgreSQL, S3, and CloudWatch.
- 90 days
- Add hosted project onboarding, broader pricing catalog support, and Amazon Bedrock cost metadata experiments.