Skip to main content

Event Log

The Event Log is the append-only audit trail of every application event that flows through Codebolt — agent starts, file changes, thread completions, job updates, and more. It sits alongside the other memory layers: the source of truth for what has happened, queryable by memory pipelines and by the dedicated panel.

Open via: Debug Tools dropdown → Event Log

What it shows

Every event in the system is captured with:

  • Event type — the category of event (e.g., agent:completed, file:updated, job:assigned)
  • Timestamp — when the event occurred
  • Payload — the full event data (expand to inspect)
  • Source — which subsystem emitted the event

View modes

ModeDescription
ListChronological stream of all events
DetailFull payload inspection for a selected event
QueryFilter and search events by type, field, or value

Event categories

Events are grouped by the subsystem that emits them:

PrefixSource
agent:*Agent lifecycle (started, completed, error)
thread:*Thread lifecycle (created, completed, message added)
job:*Job state changes (created, assigned, completed, split)
task:*Task lifecycle (created, started, completed, failed)
file:*Filesystem changes (created, updated, deleted)
git:*Git events (commit, push, pull request)
swarm:*Swarm state changes
conversation:*Conversation events
calendar:*Calendar triggers and reminders

Querying events

In Query mode, filter events by:

  • Event type — select one or more event types from the dropdown
  • Field conditions — filter on payload fields using operators: =, !=, >, <, contains, startsWith, matches, exists
  • Time range — narrow to a specific window

Conditions can be combined with AND, OR, and NOT logic.

Real-time updates

The Event Log receives push events over WebSocket. New events appear immediately as they occur. Use this alongside the Debug panels when tracing a multi-agent workflow — you can watch events from hooks, webhooks, agent completions, and job assignments arrive in sequence.

Relation to memory pipelines

The Event Log is read-only — it observes, not acts. Because it is treated as a first-class memory layer, persistent memory pipelines can log_search over it to build summaries, detect patterns, or surface activity in future context windows. See the Event log layer in Memory Layers for how it fits alongside episodic, KV, JSON, and vector stores.

Relation to Hooks and Webhooks

To react to events automatically rather than just observe them, use Hooks (for lifecycle triggers) or Webhooks (for external HTTP triggers). The Event Log helps you verify that the right events are being emitted before writing a hook or debugging why one didn't fire.