Built with
What we used, and what for.
| Tool | What we used it for |
|---|---|
| AgntID | The customer-hosted runtime and portal that make every Runtime Access Decision. We call it through the agntid Python SDK. |
| Sarvam AI | Text-to-speech for the caller voices and the on-call phone, speech-to-text for the engineer's spoken approval. English and Hindi. |
| Gemini API | The model behind live agent runs, through its OpenAI-compatible endpoint. The stage run uses a scripted agent. |
| FastMCP | Our payments and comms MCP servers that sit behind the runtime. |
| Model Context Protocol | How every agent reaches every tool, so the runtime can sit in between. |
| GitHub MCP server | GitHub's official server for issues, branches, PRs and merges. |
| PostgreSQL | The synthetic refund ledger, orders and masked card table. |
| GitHub Actions | Runs cleargate gate on every PR that touches agents or policy. |
| cryptography | Ed25519 signatures on the record, checked by verify.py. |
| OWASP LLM Top 10 (2025) | Labels on every attack in the catalogue and the coverage matrix. |
| OWASP Agentic AI threats | The ASI ids on the CI gate's attack table. |
| FastAPI | The split-screen console and the /oncall phone page. |
| Docker | Runs the AgntID runtime from its public compose file. |
| Vercel | Hosts this site. |