Browse
→ AI & Models
→ Titration
Titration
Self-hosted MCP for testing prompt changes against frozen baselines with cross-vendor judges, disagreement bands, failure-origin classification and searchable evaluation memory. Requires local Postgres and configured external judge providers; results are not a substitute for human review. Repository created September 28, 2026.
MCP unverified
Integration
| Transport | stdio |
| Auth | api-key |
| Endpoint | npx tsx /absolute/path/to/titration/server/server.ts |
| Install | git clone https://github.com/kaithoughtarchitect/titration.git && cd titration && docker compose up -d && npm install && npm run setup (Node 22.11+, Docker; configure judge credentials privately) |
Use Cases
| 01 | Compare a candidate prompt with a frozen baseline |
| 02 | Classify why an evaluation failed |
| 03 | Review judge disagreement before accepting a change |
Tags
prompt-evaluation cross-vendor baselines postgres stdio
Machine-readable: /api/servers.json
· JSON-LD schema embedded in <head>