Install AgentPM where your coding agents run. Then use the app and skill to search what happened, inspect the evidence, and carry the right context forward.
// first run
Start from the Add data dialog. The command routes captured sessions back to the right org automatically.
01
Start with the workspace that should own the captured sessions.
02
Copy the install command that is already scoped to that org.
03
Install on the laptop, dev box, or server where coding agents run.
curl -fsSL https://agentpm.dev/install.sh | bash04
Check the machine heartbeat and captured sessions in AgentPM.
// coding skill
Once AgentPM is collecting sessions, install the bundled skill so your agent can ask AgentPM for prior decisions, open work, risks, and daily progress before it starts changing code.
$ agentpm skill install
installed skill=agentpm-conversations
Pick up a thread
You
Check AgentPM for the Codex conversation where I was working on the checkout refactor. What was the last real state before that session ended?
Agent
I'll recover the prior conversation, including decisions, open tasks, and the last known blocker.
Avoid repeated debugging
You
Before debugging the auth callback again, check AgentPM for earlier Claude and Codex sessions. What did we already rule out?
Agent
I'll look for prior investigations so we do not repeat solved paths.
Recover a decision
You
Check AgentPM for why the previous agent moved checkout state out of the page component. Was that a user decision, shortcut, or bug fix?
Agent
I'll find the source conversation and recover the reasoning before editing.
Prepare a PR
You
Before I open this PR, check AgentPM for the conversations behind this branch. What was verified, skipped, and worth reviewing?
Agent
I'll build the PR context from the actual sessions, not just the diff.
// review the work
After a few sessions, start with Conversations. Search for the feature, project, or decision you care about. Then open the conversation and use each product view for a different kind of evidence.
01 / conversations
Open Conversations, search by project, repo, phrase, or feature, then jump into the session that has the context.
02 / insights
Use Insights to filter decisions, outcomes, risks, artifacts, config changes, state changes, and advice with links back to exact turns.
03 / notebook
Use Notebook when you need the raw thread: prompts, agent responses, tool calls, terminal output, edits, retries, tokens, and cost.
04 / tools
Use Tools to understand where work happened: shell commands, reads, edits, writes, Git operations, and tool-result evidence.
05 / daily digest
Use Daily Digest to see what moved forward, which projects were active, what advice surfaced, and what follow-up deserves attention.
// ready
Import existing conversations or capture new sessions as they happen. AgentPM turns agent history into decisions, outcomes, open work, and context future agents can use.