Skip to main content
Sediment is an open-source, self-hosted evidence store for coding agents. You can’t tell which agent output developers kept, which commits it reached, or whether it passed CI. Sediment records model calls, code changes, developer decisions, and check results on your infrastructure, and links them.
Evidence about one piece of agent output arrives from separate systems. A coding agent sends an Inference call, a Developer decision, and an Edit observation. Git and CI send a Push and a CI outcome. Sediment appends each one as a Fact in PostgreSQL, and the stored count rises from 0 to 5. From those Facts, Sediment derives an Attributed completion: Developer decision accepted, Edit retention score 0.92, Commit a1b2c3d, CI outcome passed. Reports and Training rows read the Derivation. Agent context reads the Facts.Evidence about one piece of agent output arrives from separate systems. A coding agent sends an Inference call, a Developer decision, and an Edit observation. Git and CI send a Push and a CI outcome. Sediment appends each one as a Fact in PostgreSQL, and the stored count rises from 0 to 5. From those Facts, Sediment derives an Attributed completion: Developer decision accepted, Edit retention score 0.92, Commit a1b2c3d, CI outcome passed. Reports and Training rows read the Derivation. Agent context reads the Facts.

Quickstart

Capture your first agent session on one machine in about five minutes.

How it works

One evidence store for evaluation, agent context, and training datasets.

Evaluate agent work

Compare model outcomes, code retention, and rework.

Reuse context

Select captured evidence for an agent continuing a task.

Deploy

Gateway capture, forge webhooks, per-machine setup, and verification.

Build training datasets

DPO, SFT, diff-SFT, Recovery, and RLVR rows.