Agents that remember
what went wrong.
The gateway does more than meter model calls. It turns opted-in session evidence and agent observations into repository-scoped insights, prompt feedback, and compact project memory that teams can inspect before agents reuse it.
Runbook snippets, recent lessons, repo conventions, and sharp edges.
Attributed insights, prompt feedback, coverage, and curated memory in one repository view.
Failures, successful fixes, commands, and code pointers distilled from real work.
Context is created by doing the work.
Eligible sessions can produce project knowledge. The useful parts are attributed, distilled, reviewed, compacted, and made available through explicit project tools or centrally controlled context policy.
- 01
Project identity
The client reports Git remotes and the gateway resolves them into one tenant-scoped repository project before knowledge is read or written.
- 02
Session evidence
Opted-in session metadata and canonical logs give the distiller evidence without mixing raw prompts into lean usage rows.
- 03
Observation tools
Agents can record useful facts, fixes, and sharp edges while they work, instead of relying on someone to write a recap later.
- 04
Insight distillation
A nightly, allowlisted pass turns completed sessions into attributed decisions, mistakes, lessons, context, and prompt reflections.
- 05
Memory compaction
The compactor folds repeated lessons into a smaller project memory file so future sessions get signal, not a growing transcript dump.
The important part is not recall. It is not repeating errors.
A good project knowledge system separates decisions, broken assumptions, recovery commands, prompt feedback, and curated memory. The gateway makes each layer visible instead of treating a transcript as an opaque source of truth.
Remember the exact repo command that fixed a broken local test harness.
Carry forward the reason an API route must stay fail-closed.
Warn the next agent before it repeats a bad deploy path.
Show where an insight came from before turning it into project memory.
Knowledge is visible, attributed, and auditable.
Teams can inspect repository memory, generated insights, prompt feedback, project identity, and distillation coverage. Memory controls remain available for explicit retrieval and injection policy.
# generated insights
Decision — Use the normalized Git remote as the stable project identity.
Mistake — Do not report unavailable knowledge as zero.
Context — Attribute accepted model requests independently of raw log upload.
Name the target environment and evidence boundary up front to reduce a second investigation pass.