Runtime context that stays in your environment.
Airgapped and on-prem by design. There is no AppMap cloud for traces.
Bring your model. Keep your evidence.

Architecture governance cannot rely on the AI grading its own homework. AppMap adds an independent evidence layer inside your environment: runtime behavior captured from the application itself, outside the AI's reasoning loop. The agent can consume it, the reviewer can see the same evidence, and the organization can govern against it. The AI explains the evidence. It does not create the evidence.
Architecture and trust
Developer environment
AppMap records and compares traces
tmp/appmap · working traces
CI
The same review runs in the pipeline
No outbound dependency on AppMap services
Repository
gold_traces/ · sanitized before commit
AppMap runs inside your environment with no outbound dependency on AppMap services. Offline activation and internal distribution are supported. There is no AppMap cloud for traces.
Trust, but verify: open-source clients and a published data spec. Developers and security reviewers can audit what AppMap records and what it sends to the agent.
How AppMap deploys
What the reviewer sees
The same runtime evidence the agent queried, generated locally and attached to the change.



From one workstation to CI
- Step 1
Start on the workstation
Teams do not need to instrument CI to begin. AppMap records from a running process, so teams start in the editor or in environments that already run: a local run, a focused test, a smoke script, an API call, or a QA environment. No full enterprise-stack build required to record the first traces.
- Step 2
Centralized ground truth of application behavior, in git
The AppMap Gold Traces skill identifies important application paths and uses existing tests to build the set. When coverage is missing, AppMap suggests a new test case. The organization can run the same review in CI when ready. The set lives in the team's repository, where changes follow the same pull-request workflow as the code. Teams begin lightweight in the editor and graduate the flows that matter into the pipeline.
Public open-source demo: FINOS Waltz