Skip to content

Observability & CLI

Every mono-agent run produces bounded local JSONL evidence. The artifacts are the on-disk record after successful recorder boundaries, not a crash-safe in-flight journal. A trace-source registry lets operator surfaces discover running agents, the mono-agent CLI operates the lifecycle, and the always-on web console provides live operation.

The framework no longer bundles a Phoenix/OTLP trace exporter or historical export command. This does not make an agent network-isolated: configured providers, channels, MCP servers, web tools, and external memory services can still use the network.

SurfaceWhat it isCoveragePage
JSONL run artifactsPer-run run-*.events.jsonl + run-*.summary.json; sensitive-looking values are redacted and retained free text is scanned for a closed set of credential shapesconfig / autoRun artifacts & traces
Trace-source registryHeartbeat manifests for local operator discoveryconfigRun artifacts & traces
Run reportsRead-only audit, metrics, and bounded prior-run evidence over local artifactscli / codeRun artifacts & traces
mono-agent CLIinit / validate / start / stop / logs / restart / web / runs / install-skillcliCLI reference
Web consolePersistent multi-agent conversations, streamed turns, and local-device attachmentscliWeb console

At start(), the recorder separately replaces an empty events file and a running summary. It applies bounded string and event limits, schedules best-effort checkpoints, and writes a terminal snapshot at finish() or fail(). Non-numeric values under sensitive-looking object keys are redacted; numeric values under matched keys are retained; retained free text is scanned for a closed set of high-confidence credential shapes.

{
"artifacts": { "dir": "./.mono-agent/artifacts" }
}

Set artifacts.dir in JSON to override the directory. The tool bloat guard also persists oversized tool output beneath the artifact root.

Each summary has a final status (succeeded, failed, cancelled, or interrupted). A run left at running by a dead prior process is reconciled to interrupted at the next startup. mono-agent runs, runs audit, and runs report read these artifacts locally without contacting a collector.

See Run artifacts & traces for the write boundary, event schema, retention, history, and metrics contracts.

The host publishes a heartbeat manifest so local operator surfaces can discover which agents are running and when a source becomes stale.

{
"traceability": {
"registryDir": "./.mono-agent/trace-sources",
"sourceId": "my-agent",
"sourceLabel": "My Agent",
"heartbeatMs": 10000,
"staleAfterMs": 30000
}
}

Configure traceability in the JSON traceability block.

First-party Phoenix/OTLP support, observability.exporters, MONO_AGENT_OBSERVABILITY_EXPORTERS, and mono-agent backfill are removed. Active legacy config fails before startup with fixed migration guidance instead of being silently ignored. An absent field, a blank environment assignment, environment [], observability: {}, or observability: { exporters: [] } is accepted only as an inert upgrade tombstone and does not enable an exporter.

There is no automatic export, replacement selection, config rewrite, artifact conversion, local deletion, or remote-trace deletion. If a final historical export is required, perform it before upgrading with the complete currently working version set already used by that consumer. The public registry does not provide a separately installable Phoenix plugin, so this documentation does not invent a package pin. See Framework simplification migration and Deprecations.

mono-agent drives the agent lifecycle from one config: init scaffolds non-destructively, validate prints a per-section report, start launches traceability plus configured channels, and stop / logs / restart operate the running instance. runs audit scans local summaries read-only and runs report aggregates local latency, cost, and failure rates.

Invoking mono-agent backfill or mono-agent help backfill now prints the removal and pre-upgrade migration pointer. The full command and flag matrix is in the CLI reference.

mono-agent runs list and mono-agent runs show <run-id> read bounded, redacted run evidence without contacting a provider. mono-agent config provides the resolved configuration view. The former terminal renderer is removed; use the web console for live operation.

mono-agent web start installs the persistent browser conversation console on macOS; mono-agent web run is the foreground cross-platform path. It auto-discovers running agents and keeps threads and in-flight work in an owner-private service store.

Terminal window
mono-agent web start
mono-agent web

A fresh install binds 127.0.0.1:5050; --host <addr> explicitly widens the listener. There is no application login, so network reachability is authority to operate the agents. See the web console guide for the complete boundary.