CONCEPT-001 · CHECKED

What is an agent run record?

An agent run record is the stored history of one execution by an AI coding agent: the inputs it received, the model it used, the tool or command steps it ran, the outputs it produced, the final status and any artifacts it left. It exists so a person can inspect what actually happened instead of trusting a chat transcript.

Audience
Reviewers of AI coding agent output
Checked
2026-11-19
Boundary
Not a correctness certificate

§ 01

Definition

An agent run record is the stored history of one execution by an AI coding agent. It captures the inputs the agent received, the model that ran, the tool or command steps it took, the outputs it produced, the final status and any artifacts it left behind.

The record is kept by a separate layer from the agent itself, so it reflects what the infrastructure observed rather than what the agent chose to report. AdaptOrch keeps run records for hosted MCP and API runs; other tools may keep similar records under names like execution receipt, session audit or trace.

§ 02

What a run record contains

  • Inputs — the task text, context and configuration the agent was given.
  • Model — which model executed the run (hosted or your own key).
  • Steps — tool calls, commands and their recorded outputs, in order.
  • Status — how the run ended: succeeded, failed, cancelled or still open.
  • Artifacts — files, logs and exports the run produced.
  • Checks — which verifications ran (if any) and their outcomes.

A completed run is not a test pass. The record shows what ran; it does not prove the change is correct.

§ 03

Why a run record matters

When a coding agent says "tests pass" or "I updated three files", a chat transcript is a claim. A run record is evidence. The reviewer can compare the claim against the recorded commands, outputs and skipped checks, and see what was never verified before deciding whether to merge.

This is the gap AdaptOrch fills: it does not write code and does not comment on diffs; it keeps the record so a person can review the agent's work. See the demo run for what a record looks like.

§ 04

Run record vs chat transcript

A transcript is what the agent said. A run record is what the infrastructure observed: tool calls, exit codes, recorded outputs and status, kept outside the agent's own narration. Transcripts can be edited or summarized away; a stored run record keeps the steps and their evidence.

§ 05

FAQ

Q.01Is a run record the same as a log?

A log is usually unstructured text. A run record is structured history: inputs, model, ordered steps, outputs, status and artifacts, stored for later inspection.

Q.02Does a run record prove the change is correct?

No. It shows what ran and which checks were available. A passing check covers only what it executed; a missing check stays unverified.

Q.03Who keeps the run record?

A separate layer from the agent. AdaptOrch keeps hosted run records so a reviewer can inspect them later instead of relying on the agent's own report.

NEXT

Try it on one agent change.

Connect your coding agent, run one change through it, and read the record next to your usual review. Starter is free.

START FREE ON STARTER

Starter is free with 50 admitted runs a month. Pro is $39 a month for 500 admitted runs and 5 workspace users. Team is $149 a month for 2,500 admitted runs and 15 workspace users. Compare plans · Start free on Starter

Connect from Claude Code, Codex CLI or Cursor · Read the AI PR review guide