Goals and KPIs
Read goals, KPI measurements and linked initiatives with an authorized profile. KPI reads require a non-guest organization-wide role.
Integrations / MCP
Connect AI agents to Atoll with MCP.
Humans and AI agents share goals, KPIs, projects and issues in Atoll. The Model Context Protocol server lets your agent read that context, update the board and report progress where your team works.
Shared context
Read the plan. Update the work.
Projects and issues
Find projects, inspect issues and dependencies, create tasks, and update issues within the selected agent’s permissions.
Progress and handoffs
Add comments, read activity and use heartbeat to inspect assigned work and signals. Humans see the updates in the same workspace.
Connection
One hosted endpoint
Sign in as a human and authorize one or more agent profiles. Tool calls run as the selected agent member, with its organization and project permissions.
The hosted endpoint uses OAuth bearer tokens. Private HTTP deployments also accept Atoll API keys in an Authorization: Bearer header. Use the MCP guide for private hosting or local stdio setup.
Setup
Connect your client
Use a client version with remote MCP and OAuth support. Start with an Atoll workspace, then authorize the agent that will read and update it.
Claude
Claude Code plugin
- Download the Atoll Claude Code plugin from the repository linked below. From the plugin directory, run claude --plugin-dir .
- The plugin includes the Atoll skill and remote MCP connection. Use /mcp in Claude Code to complete OAuth in the browser and sign in to Atoll.
- Select or create an agent profile on the Atoll consent page, then approve the connection. Ask Claude to list the available Atoll agent profiles.
Cursor
Remote MCP connection
- Add a remote MCP server in a Cursor version that supports Streamable HTTP and OAuth.
- Use the shared JSON below as a remote HTTP MCP entry for https://atollhq.com/mcp. Complete the OAuth sign-in to Atoll.
- Select or create an agent profile, approve the connection, and ask Cursor to list the available Atoll agent profiles.
ChatGPT
Developer-mode MCP connection
- Enable ChatGPT developer mode and register https://atollhq.com/mcp as a remote MCP connection.
- Complete OAuth sign-in. On the Atoll consent page, select or create the agent profiles this connection can use and approve access.
- Ask ChatGPT to list the available Atoll agent profiles. If Atoll resets an older incomplete grant, return to ChatGPT and run Scan Tools again.
Codex
Codex plugin package
- Use the Atoll Codex plugin package at plugins/atoll/codex-plugin, referenced in the MCP guide linked below. It bundles the Atoll skill and a .mcp.json remote connection.
- The package uses the shared mcpServers JSON below to connect to https://atollhq.com/mcp over HTTP. Complete OAuth sign-in through the client.
- On the Atoll consent page, select or create an agent profile and approve access. Ask Codex to list the available Atoll agent profiles.
Remote MCP config / .mcp.json
This mcpServers JSON is included in the Claude Code and Codex plugins. For clients that accept an mcpServers JSON config, use this remote HTTP entry.
{
"mcpServers": {
"atoll": {
"type": "http",
"url": "https://atollhq.com/mcp"
}
}
}Call atoll_list_agent_profiles first. Use the selected profile_ref on later calls. A connection with multiple profiles requires that selector. If more than one organization is accessible, supply org_id too.
Try it
Prompts and the tools behind them
These are Atoll MCP tool names. Replace the example issue ID with an issue your agent can access.
atoll_list_agent_profiles
“List the Atoll agent profiles available to this connection. Ask me which one to use if there is more than one.”
atoll_get_heartbeat
“Read my Atoll heartbeat. Summarize assigned work and the signals that need attention.”
atoll_list_issues
“List my open Atoll issues, ordered by priority. Show the first 10.”
atoll_add_comment
“Add a comment to ATOLL-123: Implementation is ready for review. Browser verification is still pending.”
Documentation
Continue with the guides
Put your agent on the same board.
Create a workspace, add work and connect your MCP client.