EN HR SL
Start Free
AI execution · step 3 of the loop

Your roadmap is your agent's backlog.

Connect any MCP-compatible coding agent — Claude Code, Cursor, your own tooling — to Damper's hosted MCP server. One URL, one API key, no local install. The agent reads your context, picks the highest-demand task, and ships it.

Start free MCP reference
Setup

Setup in one config file

Add the Damper MCP server to your assistant. It uses streamable HTTP transport — nothing to install locally.

Every request is authenticated with your API key in the Authorization header. Create keys under Settings → API Keys in the dashboard — they use the dmp_ prefix.
The loop

The recommended loop

Six calls cover a whole task, from reading the rules to handing it back.

  1. Load context

    get_project_context pulls architecture, conventions and critical rules into the session. Always first.

  2. Pick a task

    list_tasks returns available work sorted by weighted demand score — the roadmap decides, not the agent.

  3. Lock and read the spec

    start_task locks the task and returns its full spec, linked feedback and project rules.

  4. Do the work, leave breadcrumbs

    add_note records decisions; add_commit attaches commit hashes. Future agents see everything.

  5. Complete with evidence

    complete_task requires confirmation for every checklist item. No evidence, no ship.

  6. Or hand off cleanly

    abandon_task releases the lock with a written summary so the next agent starts where you left off.

Ship announcements included. When an agent completes a public task, Damper auto-drafts a changelog entry and notifies everyone who voted for it.
Locking

Task locking, built in

A task started by one agent is locked against all others — across teammates, machines and agents. Locks carry the owner and timestamp, and every conflict surfaces a clear handoff path instead of silent double work.

For projects where docs alone aren't enough, the context graph adds typed relationships — with authority and confidence on every edge — so agents resolve exactly which rules govern a change.

Context graph
Tools

What agents can do

100+ tools across tasks, context, pages, feedback and releases. The highlights:

01

Task lifecycle

Pick, lock, track and finish work — with dependencies, PRs and a full audit trail.

02

Context & graph

The docs layer your agents actually read — and the graph that maps it.

03

Pages, feedback & releases

Agents can read the workspace, answer customers and publish the release.

04

Also on board

Subtasks · dependencies · task review · checklists · code templates · module registry · settings · self-assessments

See all 100+ tools

Give your agent a real backlog.

Every task your agent finishes closes the loop with the customers who asked for it.

Start free MCP reference
The product map

Where this fits in the loop.

01

Collect

Catch feedback where it happens — in your app, in Slack, on a call — and land it in one clean inbox.

Feedback widget Feedback inbox
02

Prioritize

Rank requests by who is asking. Votes are weighted by plan, so paying customers move the roadmap.

Weighted voting
03

Build

Hand the top task to your AI coding agent with the rules that govern it — and get proof when it is done.

AI execution Context graph Review & proof
04

Ship

Shipped work becomes a changelog entry, and everyone who asked for it hears first.

Changelog
+

Around the loop

Docs, uptime, errors and team access — the product ops that sit next to the loop.

Pages Status pages Error logs Projects & teams