Loading your next step…
Loading your next step…
Bring your own agent
Use your own AI agent to clarify what matters, choose a next step, and save your progress.
Requires an agent with HTTP tools or a terminal. No repository checkout needed.
Let your agent read a chosen goal, its specification, affirmations, progress, and latest handoff. Discuss what that aim means to you.
The guide helps your agent clarify your desired outcome, use reading, reflection, and writing, and agree on a manageable action.
With write access, save agreed steps, evidence, and handoffs in the same workspace you see in the app. A fresh session can pick up where you left off.
On macOS/Linux, restrict the connection file and use its absolute path. From the extracted skill directory, with Node 22 or later:
chmod 600 /private/path/achieve-aims-connection.json export AIM_CONNECTION_FILE="/private/path/achieve-aims-connection.json" node scripts/aims.mjs list node scripts/aims.mjs ideas # if idea sharing is enabled node scripts/aims.mjs read GOAL_ID
Replace the path and goal ID with your own. On Windows, use private file permissions and set the same environment variable in your shell or host settings. The client prints private goal context; keep its output in your authorized session.
After connecting, paste this into your agent. It contains instructions, never your key.
The integration uses ordinary authenticated HTTPS and JSON. A terminal agent can run the dependency-free client; an HTTP tool host can use the OpenAPI contract directly. We test the downloadable client from a clean folder against the real API handlers, including retries, resuming work, and access restrictions.
| Host | What it needs |
|---|---|
| Terminal agent | The skill folder, Node 22+, outbound HTTPS, and private connection configuration. |
| OpenClaw | Place the folder in your workspace’s skills directory and enable the execution tool. Host setup |
| Hermes | Place the full folder in its skills root and provide the credential to its terminal environment. Host setup |
| Grok in a tool-enabled host | Use the template below with a terminal or a host that executes the supplied function definitions. |
| Chat without execution tools | Can discuss a plan you provide. Cannot read or save private goals from a prompt alone. |
OpenClaw and Hermes setup follows their documentation; we have not certified every host or version. There is no hosted MCP endpoint. The public HTTP and terminal clients work independently of the repository’s optional local MCP bridge.
The connection accesses selected unencrypted goals and their workspaces. You can separately share unlinked ideas and ideas attached to those goals for reading. Other goals and their linked ideas, encrypted content, scratchpads, billing, and account settings stay outside its scope. Reading public instructions grants no account access.
To add idea access to an existing agent, create a new connection with idea sharing enabled and replace its key in the host’s secure configuration. Download the current skill ZIP or reload the function definitions so the agent knows how to read ideas. Revoke the old connection after the replacement works.
Writing covers work items, notes, evidence, workspace criteria, and handoffs. With separate goal proposal permission, your agent can draft new goals or changes to titles, specifications, status and order. Open its review link to see every change and approve the batch once. Nothing changes until you approve. New goals are shared with the agent only if you choose that during review. Affirmation edits and daily practice logging still happen in the app.
Data you share with an external agent is also handled by its host and model provider. Choose a host you trust and share only the goals it needs. Read our privacy policy.
The Grok template includes instructions and tool mapping; it is not a hosted bot. The skill checksum manifest lets you verify the ZIP.
HTTP: list with GET /api/v1/goals, read shared ideas with GET /api/v1/ideas, read GET /api/v1/goals/GOAL_ID/workspace, and save a command with PATCH at the workspace URL. Supply the current workspace version and a UUID operationId. Command examples, retry rules, and limits.
Work through a fresh set of aims with your coach, then review one draft containing up to 20 changes. Create goals, sharpen their specifications, defer an old aim, or reorder your priorities.
At Agent connections, enable “Allow this agent to propose goal changes for my approval.” This is separate from workspace writing. Existing connections need a replacement key with this permission. Reload the current skill or Grok template after updating.
Your agent gives you a private review link. You see before and after, then approve all changes or reject the draft. Drafts expire after 24 hours; if your goals changed meanwhile, your coach needs to prepare a fresh one. New goals count toward your plan limit.