Planning work for Claude Code
The short answer
This page is for when you run Claude Code and need a plan around it: what it works on next, who picks up when a session ends and how you know a job is done. Each question below gets a one-line answer and a link to the guide that goes deep.
You run Claude Code, and it does much of the building. What it works on next, what it remembers and how you know a job is done are calls about the plan, and the plan should not live inside one session.
What Claude Code can do, how to install it and its commands are in Anthropic's own docs. This page covers planning and coordinating the work.
Why plan outside the session?
A session ends. Your priorities, your customers and what you decided last week should not end with it. Keep them in one place every agent can read, and each new session starts from the plan, not from a blank page.
Where Annsa fits
Annsa works like an operating system for small business run with AI agents: your direction, goals, projects, work and customers, in one plan. Claude Code, Cursor, Codex and Grok read that plan over MCP, claim one job each and hand over with a note. Annsa runs its own work this way. In Cat's words, "So September became about getting us all working from the same direction." Read September: getting our agents in a row for how that went.
The guides below work without Annsa too. Start with the question you have today.
Questions people ask.
Does Claude Code plan its own work?
What is the smallest plan that works for one founder and one agent?
Do I need a tool, or are files enough?
Where do I find how to install or use Claude Code itself?
A guide to planning work for Claude Code.
- 01
How do I hand work from one Claude Code session to the next?
Write a short note when a session stops: what is done, what is next and what was decided. The next agent reads it and carries on.
- 02
How do I run several agents without them colliding?
Keep one ranked list, write one plan per job and give each job one holder. Give each agent its own branch.
- 03
Should I use Claude Code agent teams?
Only for splitting one big task inside Claude Code. They are experimental, they do not choose which tasks are worth doing, and Cursor, Codex and Grok cannot join them.
- 04
How do I keep specs current while agents build?
Write the spec before the job starts and keep it as the source of truth. When the work changes, the spec changes, and you approve the update.
- 05
What should Claude Code remember between sessions?
Standing rules in CLAUDE.md or AGENTS.md, and nothing that changes weekly. Priorities and decisions belong in the plan.
- 06
Where do decisions and context live?
In the plan, written down once so every agent reads it. Keep each decision with its reason, so nobody answers the same question twice.
- 07
How do I keep a long task on track?
Cut it into jobs, give each a Done When line and have the agent leave a handoff note at every stop. Check each line before you call it done.
- 08
Which project management tool works with Cursor and Claude Code?
One your agents can read and write over MCP: the ranked list, a spec per job and a claim on each.
- 09
How do I connect Claude Code to my plan over MCP?
Add the server once and Claude Code reads the ranked list, claims a job and records what it shipped.
The source of direction for agent teams.
Begin with one agent and the work that matters next.

