Start here

Give your agent memory

Your coding agent forgets everything when the session ends. pensieve fixes that: install the local binary, wire it to your agent, and it recalls durable, graph-aware memory into every prompt. One binary, zero infra — about five minutes.

1. Install — zero infra

One command. Installs the pensieve CLI and a local engine backed by embedded SQLite + local files under ~/.pensieve — no Postgres, no Docker, no sudo.

pensieve ~ install

$curl -fsSL https://raw.githubusercontent.com/shakedaskayo/pensieve/main/install.sh | bash

2. Wire your agent

Three paths, same engine underneath — they compose, so you can start with one and add the others later.

Claude Code · automaticThe pluginHooks capture every turn and inject the most relevant memories into every prompt — no tool call. Plus /pensieve-recall, /pensieve-remember, /pensieve-ask. The "it just remembers" path.pensieve install-pluginSet up the plugin →Any agent · MCP or CLIMCP / CLI + skillCursor, Windsurf, Aider, Continue — wire the full toolset over stdio MCP, or install a skill that teaches a shell-tool agent when to recall and remember.pensieve setup claude-codeConnect any agent →

3. Remember, recall

Memory round-trips through the same engine the plugin, the CLI, and the web UI all use:

pensieve remember "payments-svc deploys behind the Aurora gateway; error budget is 0.1%."
pensieve recall   "how do we deploy payments and what's the error budget?"
# → returns the memory, scored by vector + keyword + graph

Recall is graph-aware hybrid retrieval — semantic + keyword, fused with Reciprocal Rank Fusion, then expanded over the memory graph. It's durable across sessions and machines. With the plugin, you rarely call recall yourself — the right memories are injected into each prompt automatically.

That's the whole hook

Install → wire → remember/recall. Your agent now carries context across every session. Everything below is pensieve growing from there.

Then it grows

Memory is the front door. The same engine and the same graph keep going:

Live data · point logs, traces, and code at it; query in KQL/SQLHow it all works · the memory → data → graph model, end to endDreaming · background agent that keeps memory sharpSync · keep memory coherent across machines and your team

Where to go next

• The mental model, end to end: How pensieve works.
• The full connect surface (plugin · CLI · MCP): Connect your agent.
• How recall, the graph, and bi-temporal validity work: Agentic Memory.
• Run the whole engine and query live data: Five-minute start.