Track

Track durable work on Pegboards

Pegboards organize plans, tasks, and notes that an authorized AI reads and writes through the Pegs MCP surface. The app is the review surface: it groups work by lifecycle and shows each peg's current sections, status, claimant, and last-updated time.

Last reviewed July 14, 2026

In the app Track → Pegboards

Problems this solves

Start with the recurring problem, then use this feature when its boundary matches the outcome you need.

Problem

Continue complex work across AI sessions

A long project outlives one context window, so the next session loses the plan, accepted decisions, and current handoff.

How this helps

Store specs, plans, tasks, decisions, memories, comments, and a session handoff as durable Pegs instead of leaving them in chat.

Problem

Coordinate agents without duplicating work

Several participants can pick up the same task or proceed before a dependency is resolved.

How this helps

Use claims, revision checks, blocker edges, ready-work filtering, and explicit status transitions to coordinate work.

Problem

Keep implementation tied to the reason behind it

A flat task list records what to do but loses the parent plan, acceptance criteria, and decision that justified the change.

How this helps

Use a structured methodology to connect tasks to specs, plans, phases, decisions, and completion evidence.

Open or create a board

Use Main board when it is present. New workspaces receive it; older workspaces may not.
If you need another board, choose New pegboard.
Name the workstream and select Simple or Superpowers methodology.
Create the board, then open it to confirm the name and methodology.
Authorize an AI client for the board before asking it to add work.

Methodology is permanent after creation because it defines the peg lifecycle and structure. Use Simple for a lighter task flow; choose Superpowers only when the work needs its additional structured process.

Choose the methodology deliberately

Simple: Task-only boards with ready → in_progress → done.
Superpowers: Specs, plans, tasks, decisions, and memories with enforced parent, section, phase, and status rules.

A methodology is validation, not just display labels. Ask the AI to read the board's methodology schema before creating structured work instead of assuming every board accepts the same types or sections.

Connect the AI to Pegs

Pegboards use the same hosted MCP endpoint as Docs, Crews, and Dashboards: https://app.solidactions.com/mcp. Copy that URL into the AI client you want to use and complete its browser authorization. Pegboard access is separate from a workflow API key and from workflow provider Connections.

If your client is already connected to the hosted MCP service, approve the Track/Pegs product area and intended board scope. You do not install a Pegs server or local MCP package.

Add and update work

The board detail is read-only in the current app.

Create and delete whole boards in the UI. Ask your connected AI to add, claim, update, complete, or supersede individual pegs. This keeps structured board mutations on the Pegs tool surface.

Example request:

On the Website launch pegboard, add a task to verify every public docs route.
Include acceptance criteria for canonical URLs, internal links, and the production build.
Do not claim the task yet.

Use the session loop

Start with peg_session action=prime to load the methodology, in-flight claims, handoff, and ready work.
Inspect a peg with peg_get, then claim it with its current revision.
Move it to the methodology's active status and leave comments as the readable changelog.
When blocked, add a blocks edge to the blocking peg, explain it in a comment, return the task to ready, and release the claim. The dependency keeps it out of ready work until the blocker closes.
Close finished work with a final comment, or write a session handoff before context runs short.
Finish with peg_session action=audit_close and resolve any dangling work it reports.

Review the board

In progress: Active work, including the current claimant when one exists.
To do: Open work that has not reached an active state.
Done: Terminal work the process considers complete.
Superseded: Work replaced by a newer decision or item rather than completed normally.

Expand a row to inspect its structured sections. A sparse item may show no detail yet; ask the AI to add the missing outcome, context, or acceptance criteria instead of relying on a long title.

Use claims deliberately

A claim communicates who is actively working an item. It is not a permission lock. Ask the AI to check the current state before claiming, avoid taking work already owned by another participant, and release or complete the item when work stops.

Permissions and sharing

Track Read: View boards and current peg details in the app.
Track Use: Authorize one board for hosted MCP and operate its pegs with AI tools.
Track Build: Create and permanently delete whole Pegboards.

Boards are shared through workspace membership, not public links. Each OAuth connection selects one board; an API key may also be scoped to one board. Claims coordinate work but do not grant access.

Delete only when the board is no longer needed

Deleting a Pegboard permanently removes its pegs, edges, comments, memories, and audit history; it cannot be undone. Active MCP OAuth bindings or board-scoped API keys block deletion, so revoke those connections first. Supersede or close an individual peg when the board itself should remain.

Individual pegs and forgotten memories do not have a normal restore workflow. Revision numbers prevent blind concurrent overwrites, but they are not a user-visible body history or rollback system; use comments and session handoffs to preserve the change narrative.

Concurrency and practical limits

Peg writes require the latest expected_revision. On a revision mismatch, fetch the peg again, review the intervening change, and reapply only what is still needed. Create, update, and close batches accept up to 100 pegs; list operations paginate and may cap results. Keep large programs hierarchical instead of treating one board response as an unlimited export.

Next: turn tracked outcomes into measurable dashboards.