Work on Your Goals with Your Own AI Agent

Begin with a goal that belongs to you

Imagine you have a goal called “Write a collection of short stories.” You still care about it, but the last few weeks have produced more guilt than writing. A form can capture the target. A conversation may be a better place to discover why you have stopped.

An authorized agent can read your saved specification and handoff, then ask a focused question: What would making progress on this goal give you right now? You might realize that the immediate aim is to recover a creative practice, rather than finish a publication-ready manuscript this month.

The agent should help you express that distinction in your own words. It should not decide your ideals for you, turn every passing thought into a commitment, or pretend to know private information you have not shared.

Connect once, then work in the conversation

Start by creating or choosing an unencrypted goal in Achieve Aims. Open Agent connections, select the goals to share, and choose an access level. Read-only is enough for a conversation grounded in your saved context. Enable writing when you want the agent to maintain agreed steps and handoffs.

Download the private connection file when the key is shown and keep it in your agent host's private configuration. Then download the complete coaching skill. The setup guide explains the terminal client, HTTP API, and host requirements.

The distinction matters: the skill teaches a workflow; the scoped connection provides access. Giving an agent a public URL does not give it access to your account. A host without execution tools can discuss a plan, but cannot read or save your private goals.

A short session, with a useful result

Here is an illustrative exchange, not a transcript or a promised outcome:

You: “I keep putting off my writing goal. Can we make it smaller?”

Agent: “Your saved goal is a short-story collection. The last handoff says choosing an idea felt difficult. Would a small creative session be useful this week, or has the larger goal changed?”

You: “The larger goal still matters. I just want to finish one outline.”

Agent: “What is a realistic first session?”

You: “Twenty minutes tomorrow morning. Help me keep track of that.”

With write access, the agent can save a work item such as “Outline one story for twenty minutes,” with success criteria you have agreed on. It can record your next action and the question to revisit later. The workspace keeps that decision visible in the app too.

This API writes shared work items, criteria, notes, evidence, and handoffs. Parent goal creation, specification changes, affirmation edits, and daily practice logging still happen in the app. Your agent can help prepare those changes without claiming to have saved them.

Keep the method in the conversation

The downloadable guide describes Achieve Aims' adaptation of Raymond Hull's How to Get What You Want: clarify a desired outcome in writing, return attention to it, and act on useful ideas.

In a conversation, that can mean reading your aim, thinking through a practical next step, and writing it in your own words. If you like visualization, use it. If you prefer plain reflection, use that. An affirmation can remind you of your intention; it is not proof that the result already exists.

The point is to connect reflection to an action you choose. The guide does not promise that thought alone changes circumstances, reproduce the book's exercises, or prescribe a fixed script for every person.

Return without starting over

At your next session, the agent reads the current workspace and handoff. It can ask how the outline went and help you decide what to change.

If you report that you completed the session, it can save that as user-reported evidence. Achieve Aims requires an evidence note and completed dependencies before a work item can be reported done. That makes the claim reviewable; it does not independently verify it or complete the parent goal.

A useful handoff might say: “User completed one outline and wants to try drafting the opening scene. Next: choose a writing window for Saturday.” A fresh authorized session can continue from that record, even if the earlier chat is unavailable.

Share deliberately

The agent can access only the selected unencrypted goals and their workspaces. It cannot use that connection to read unrelated scratchpads, encrypted goals, billing, or account settings. You can revoke the key, and it expires by default after thirty days.

Once you share goal context with an external agent, its host and model provider also handle that data under their own policies. Choose the host and goals accordingly. You can keep using the app directly whenever you prefer.

Connect your own agent, or read why conversation can suit some goal work better than forms.