Skip to content

[Feature]: [Feature]: Session Audit Timeline — implement the "Auditable" pillar with a real-time action log UI #1573

@shaikh-amer

Description

@shaikh-amer

Summary

OpenWork's README lists "Auditable: show what happened, when, and why" as a
core pillar. but there is currently no UI that implements this. This issue
proposes building it.

Problem / goal

When an agent completes a session, users have no way to review:

  • What tools were called and in what order
  • What files were read or written
  • What shell commands were executed
  • When each action happened
  • Whether each action succeeded or failed

This matters for trust, debugging, and enterprise compliance. Right now
everything happens in a black box.

Proposed solution

A Session Audit Timeline panel in the OpenWork desktop app that:

  • Subscribes to the existing SSE /event stream (already available)
  • Displays every agent action as a chronological timeline entry
  • Shows: timestamp, tool name, input summary, output summary, status
  • Includes a one-click "Export Session Log" button (downloads JSON)
  • Works alongside the existing session composer. no new routes needed

Primary user(s)

  • Bob (IT / power user)
  • Susan (non-technical)
  • Other team roles

OpenCode primitive alignment

Thin UI layer over the existing SSE event stream. No new backend needed.
Pure frontend work consuming data that already exists.

Alignment with VISION/PRINCIPLES/PRODUCT

This directly implements the stated but unbuilt "Auditable" pillar from
the README. It makes agent behavior transparent and reviewable without
changing any agent behavior itself.

Testability

  1. Start a session, run a prompt that triggers tool calls
  2. Open the Audit Timeline panel
  3. Verify each tool call appears as a timeline entry with timestamp
  4. Click "Export Session Log" — verify JSON downloads correctly
  5. Verify existing compose/send flow is unchanged

Ready to build it yourself?

Yes

Additional context

I'm the author of agent-work-proof,
an open standard for AI agent work history and cryptographic proof of
completed tasks. Auditability is the exact problem I've been building
solutions for. This PR is a natural extension of that work into the
OpenWork UI.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions