Grounding coding agents in customer reality.

Agent product memory is a persistent, queryable store of what your customers actually asked for — Slack threads, tickets, sales calls, decisions — structured so a coding agent can ground its work in it. Your agent already knows your code; product memory gives it your customers.

The reliable way to go from a customer issue to a merged PR with a coding agent is a grounded plan file: retrieve the real customer signal behind the issue, check it against prior decisions, write PLAN.md with cited evidence and explicit non-goals, then have the agent execute that file verbatim.

PLAN.md is a one-page markdown spec with YAML frontmatter and seven fixed sections — Goal, Evidence, Constraints, Non-goals, Acceptance criteria, Implementation steps, Test plan. It works because agents execute fenced, cited, self-contained instructions far more faithfully than prose briefs.

Coding agents over-build because prompts state the goal but not the boundary. The fix is structural, not better prompting: give the agent a plan with an explicit Non-goals fence, cited acceptance criteria, and a hard gate on anything that contradicts a prior decision.

Run npx --yes relayai-cli@latest doctor, sign in, enter the exact committed Git root, run init, and require brain status --check to report READY, fresh, and PASS. Then add customer signal and generate PLAN.md. Login installs the Claude skill; init separately writes the repository's managed AGENTS.md section.

Conversation memory (Mem0, Zep) helps an agent remember its chats with you. Product memory helps a coding agent know what your customers asked for and what your team already decided. They solve different problems: one makes agents more personal, the other makes them ship the right thing.

Founders lose their sharpest product insights because they arrive at the wrong time: 2:47 AM notes, hallway remarks, one line in a sales call. The fix isn't discipline, it's a capture path with zero friction that feeds directly into what gets built: pipe the note into product memory, and it resurfaces cited in the next plan.