MCP Server · v0.1.0

Your AI agent,
inside Dhamark.

Connect Claude Code, Cursor, or Claude Desktop to Dhamark. Assign a ticket to the AI, and it picks it up, implements it, opens a PR, and marks it ready for QA, with no manual hand-off.

Get started →View tools
How it works
1
Assign to Dhamark
In the portal, set any ticket's assignee to "Dhamark" and status to "Todo". The MCP server will surface it.
2
Agent picks it up
The AI calls list_assigned_tickets → get_ticket → start_work, branches off suggested_branch, and implements the spec.
3
Auto-handed back
After gh pr create, the agent calls link_pull_request then mark_ready_for_qa, and the ticket goes back to you for review.
Setup
1
Generate an API key

In the Dhamark portal go to Settings → API Keys → Generate key. Copy the dmk_… key. It's shown only once.

Open Settings → API Keys →
2
Configure your AI host

Choose your AI coding host. The MCP server runs via npx, so no global install is needed.

Claude Code
Cursor
Claude Desktop

Run this once in your terminal:

claude mcp add dhamark -- npx -y @dhamark/mcp

Then set the env vars via /mcp or add them to .claude/settings.json.

▸ Show config for Cursor / Claude Desktop
Cursor

Create or update .cursor/mcp.json in your project root (or ~/.cursor/mcp.json globally):

{
  "mcpServers": {
    "dhamark": {
      "command": "npx",
      "args": ["-y", "@dhamark/mcp"],
      "env": {
        "DHAMARK_API_URL": "https://app.dhamark.com",
        "DHAMARK_API_KEY": "dmk_your_key_here"
      }
    }
  }
}
Claude Desktop

Edit claude_desktop_config.json (found in Claude Desktop settings):

{
  "mcpServers": {
    "dhamark": {
      "command": "npx",
      "args": ["-y", "@dhamark/mcp"],
      "env": {
        "DHAMARK_API_URL": "https://app.dhamark.com",
        "DHAMARK_API_KEY": "dmk_your_key_here"
      }
    }
  }
}
VariableRequiredDescription
DHAMARK_API_KEYyesPersonal API key starting with dmk_…
DHAMARK_API_URLnoBase URL of the Dhamark portal (defaults to http://localhost:3000)
3
Give your agent a system prompt

Paste this into your agent's system prompt or project instructions:

Use the dhamark MCP server. Call list_assigned_tickets to find todo work, get_ticket for full context, then start_work. Create a branch using the ticket's suggested_branch, implement the acceptance criteria and edge cases, open a PR with gh pr create, call link_pull_request with the PR URL, and finally mark_ready_for_qa.

The agent will then pick up any ticket assigned to Dhamark and work through it end-to-end.

Tools reference
list_assigned_ticketsread

Return all tickets assigned to Dhamark in a given status (defaults to "todo"). Start here to pick up work.

statusstring?Filter by status, e.g. "todo", "in_progress"
get_ticketread

Fetch full ticket context: description (markdown with acceptance criteria & edge cases), screenshot URL, spec, labels, and the suggested_branch the agent must use.

idstringTicket ID
start_workwrite

Move a ticket to "in_progress". Call this right after picking it up and before writing code.

idstringTicket ID
add_ticket_commentwrite

Post a progress note or update on the ticket. Markdown supported.

idstringTicket ID
bodystringComment body (markdown)
link_pull_requestwrite

Record the GitHub PR URL on the ticket after running `gh pr create`.

idstringTicket ID
urlstringPull request URL
titlestring?Optional PR title or identifier
mark_ready_for_qawrite

Set status to "qa", reassign the ticket back to the delegator, and post a "Ready for review" comment. Call this last, after the PR is linked.

idstringTicket ID
notestring?Optional extra note for the review comment
Notes
  • —stdout is reserved for the MCP JSON-RPC protocol, so all diagnostic logs go to stderr.
  • —API keys are stored as SHA-256 hashes only. To revoke, go to Settings → API Keys.
  • —The MCP server never touches git or GitHub directly. The host AI agent runs those commands with its own tools.
  • —Set DHAMARK_API_URL to your deployed portal. Leave unset only for local dev where the portal runs on http://localhost:3000.