Logline

Your app's logs, traces, and LLM calls, live inside VS Code, and ready for your AI agent.

Run your app the way you already do: in a terminal, with F5, as a task, in Docker Compose, or instrumented with OpenTelemetry. Logline puts everything it says in one searchable panel next to your code, lines production logs up with your local run, and lets Copilot, Claude Code, or Codex read it all, redacted.

See what your app is doing

• Capture anything: terminals, F5 debug sessions in any language, tasks, log files, Docker Compose and Kubernetes services, and OpenTelemetry logs, traces, and metrics, with no collector.
• Search like you think: level:error service:api, status:5xx, durationMs:>200, last:15m, or changed:true for logs from code you changed since the last commit.
• Read any format: JSON, logfmt, plain text, Pino, Winston, Log4j2, and ECS, with stack traces kept together.
• Back to the code: every log statement shows live hit counts, errors, and p95 durations above it, and Break here stops the debugger the next time it logs.

Find the problem

• Analyze shows error rate and latency over time, error groups, status codes, and the patterns behind the noise. Click anything to see its logs.
• Traces follow one request across services in a waterfall, with its logs on the spans.
• AI calls list every LLM call your app makes, from OpenTelemetry GenAI spans: model, tokens, cache share, latency, and the full conversation with each tool call.
• Log doctor flags statements that log secrets or personal data, swallow exceptions, or are too noisy, with quick fixes.

Debug production next to local

• Connections search and tail Datadog, Postgres, MySQL and Aurora (with AWS IAM sign-in), MongoDB, and SQLite logs, read-only, beside your local runs. Add one in a form that tests it and previews its log table first; Edit changes it later.
• Compare lines production errors up with your newest local runs: is it reproduced, and is your fix working? Errors from code you changed are tagged.
• Metrics chart a connection's queries; drag across a spike to fetch the logs from that window.

Hand it to your agent

Share with agent gives Copilot, Claude Code, Codex, or the agent in Cursor, Windsurf, or Kiro read-only access to what you choose: everything, or specific runs. Logs are always redacted, agents can't run anything, and Stop sharing cuts them off at once. Allow an agent to fetch from a connection, and it can pull production logs and compare them with your run, then verify its own fix.

Get started

1. Install Logline in VS Code 1.99+ from the Marketplace. In Cursor, Windsurf, or Kiro, install it from Open VSX.
2. Open the Logs panel. Click Capture off in its status bar so it reads Capture on, then run your app in a terminal as usual. Or press F5: debug output appears automatically.
3. Type level:error in the search box and press Enter. Click an event to see everything about it.

New to it? Logline offers a one-minute tour the first time, which More actions → Take the tour replays, and More actions → Help opens the guide. To try it without your own app, import samples/demo-logs.jsonl with Logline: Import Logs, or run the checkout sample.

Everything else

Logline keeps everything on your machine. Nothing leaves it unless you share it with an agent or query a connection you added.

Learn more

• Usage reference: every feature, setting, and query.
• Changelog: what changed in each release.
• Contributing and architecture: build, test, and send a pull request.

License

MIT