@postnesia/hooks

Session bootstrap hook for Postnesia. Loads core memories and high-importance L1 summaries from the database and injects them into the agent's context at session start.

How it works

At session start the hook queries the database for:

• All core memories (always loaded, never decay)
• Working memory — non-core memories with importance >= 3

Output is written to stdout in a structured block:

--- POSTNESIA MEMORY CONTEXT ---

[CORE MEMORIES]
#1 [technical] How Postnesia is used...

[WORKING MEMORY — L1 summaries, importance >= 3]
#4 [lesson] (imp:4) Learned that...

--- END MEMORY CONTEXT ---

Claude Code injects stdout from SessionStart hooks into the conversation, so the agent sees this context before the first user message.

If the database does not exist yet the hook exits silently (no error).

Usage

Claude Code (SessionStart hook)

Add to .claude/settings.json:

{
  "hooks": {
    "SessionStart": [
      {
        "matcher": "",
        "hooks": [
          {
            "type": "command",
            "command": "npx postnesia-claude"
          }
        ]
      }
    ]
  }
}

Programmatic (handler export)

The ./handler export provides a handler that can be wired into other agent frameworks. It populates the BOOTSTRAP.md bootstrap file slot with the formatted memory context:

import handler from '@postnesia/hooks/handler';

Environment Variables

Inherits from @postnesia/db — requires DATABASE_URL to be set.