THE MEMORY LAYER YOUR CODING AGENT SHOULD HAVE HAD FROM DAY ONE. CAPTURE EVERY SESSION. RECALL IN MILLISECONDS. RUN ANYWHERE.
THREE LAYERS. #
NO FRAMEWORK TAX.
BUILT ON THE iii ENGINE — EVERY MEMORY OPERATION IS A WORKER, A FUNCTION, OR A TRIGGER. NO REDIS. NO KAFKA. NO POSTGRES. THE ENTIRE RUNTIME IS ONE PROCESS.
HOOKS
12 AUTO-CAPTURE HOOKS PIPED INTO EVERY CODING AGENT. EVERY TOOL CALL, EVERY PROMPT, EVERY STOP BECOMES A COMPRESSED OBSERVATION.
RECALL
TRIPLE-STREAM RETRIEVAL — BM25 + VECTOR + KNOWLEDGE GRAPH. RERANKED ON DEVICE. P50 UNDER 20MS ON A LAPTOP.
CONSOLIDATE
HOURLY SWEEPS COMPRESS RAW OBSERVATIONS INTO SEMANTIC MEMORIES. DUPLICATES MERGED. STALE ROWS DECAYED. AUDIT ROW EMITTED EVERY DELETE.
TWELVE THINGS YOU DID NOT WANT TO BUILD. #
AGENTMEMORY IS NOT A LIBRARY OR A VECTOR STORE. IT'S A COMPLETE MEMORY RUNTIME — CAPTURE, RECALL, CONSOLIDATE, OBSERVE, FEDERATE.
- 12AUTO-HOOKS
CAPTURE EVERYTHING
Every PreToolUse, PostToolUse, SessionStart, Stop, and the rest fire into the memory pipeline without a line of glue code. Install the plugin, done.
- 53MCP TOOLS
NATIVE MCP SURFACE
memory_save, memory_recall, memory_smart_search, memory_sessions, governance, audit, export — full surface behind a single MCP server.
- 126REST ENDPOINTS
HTTP FIRST
Every MCP tool has a REST twin under /agentmemory/*. Curl it. Fetch it from the browser. Proxy it from your own agent.
- BM25+ VECTOR + GRAPH
TRIPLE-STREAM RECALL
Hybrid retrieval pipes lexical, semantic, and relational scores through an on-device reranker. 95.2% R@5 on LongMemEval-S.
- AUTOCONSOLIDATION
RAW → SEMANTIC
Hourly sweep compresses observations into semantic memories, merges duplicates, decays stale rows with retention scoring, and emits a batched audit row.
- ∞REPLAY
JSONL SESSION IMPORT
Point agentmemory at a Claude Code JSONL transcript and it rehydrates the full session — observations, tool uses, timeline — into the store.
- GRAPHEXTRACTION
KNOWLEDGE GRAPH
Entities and relations extracted on compress. Query with /agentmemory/graph. Visualize in the viewer. Temporal edges supported.
- MESHFEDERATION
PEER-TO-PEER SYNC
Register another agentmemory node, push / pull memories over authenticated HTTPS. Bearer-token required; no silent syncs.
- MDOBSIDIAN EXPORT
YOUR NOTES, HYDRATED
Mirror memories to a sandboxed vault directory. Frontmatter-tagged markdown, ready for Obsidian's graph view.
- 5LLM PROVIDERS
BYO MODEL
Claude subscription (default, zero config), Anthropic API, Gemini, MiniMax, OpenRouter. Detected from env.
- OTELOBSERVABILITY
TRACES + LOGS
iii-observability worker on by default. Exporter: memory for local, OTLP for Jaeger / Honeycomb / Tempo. Every operation produces a span.
- 0EXTERNAL DBs
ONE PROCESS
Runs as a single Node process. No Redis, Kafka, Postgres, Qdrant, Neo4j. State lives on disk as JSON. That's the whole stack.
TWO UIs. #
ONE MEMORY RUNTIME.
AGENTMEMORY SHIPS A REAL-TIME VIEWER FOR YOUR MEMORIES AND AN ENGINE-LEVEL CONSOLE FOR EVERY FUNCTION, TRIGGER, AND OTEL SPAN. BOTH ARE FIRST-CLASS — INSTALLED INLINE BY THE CLI ON FIRST RUN.
SHIP-WITH VIEWER · PORT 3113
The agentmemory server auto-starts a real-time viewer on port 3113. No install, no config. Everything the server sees, the viewer shows.
- LIVE OBSERVATION STREAM · EVERY HOOK AS IT FIRES
- SESSION EXPLORER · REPLAY ANY PAST SESSION
- MEMORY BROWSER · FILTER BY PROJECT / TYPE / CONFIDENCE
- KNOWLEDGE GRAPH VISUALIZATION · FORCE-DIRECTED
- HEALTH DASHBOARD · HEAP / RSS / EVENT LOOP LAG
$ open http://localhost:3113
MEMORY THAT TYPES BACK. #
VS. THE FIELD. #
NUMBERS STRAIGHT FROM THE LONGMEMEVAL-S BENCHMARK AND EACH PROJECT'S OWN DOCS. SHIP WHAT YOU WANT — WE JUST PICKED THE ONE WITH RECEIPTS.
IN THE WILD. #
Verbatim from the Product Hunt launch thread. Each card links back to the source comment.
Backfilled a month of Cursor transcripts
I backfilled agent memory on my past month's Cursor agent transcripts. It was surprisingly accurate. Picked up on things that I moved away from.
Been using it for 2 weeks, and I definitely see improvements.
Tackles one of the biggest pain points with coding agents: losing useful project context across sessions without bloating the context window.
The focus on making memory actually useful for agents instead of just storing context endlessly.
Memory often becomes just more noise over time. Agentmemory feels more intentional compared to a lot of tools in this space.
Tried it briefly — feels clean and easy to get started with.
SEVEN NATIVE PLUGINS. #
REST MCP-NATIVE.
NATIVE PLUGINS FOR CLAUDE CODE, COPILOT CLI, CODEX CLI, OPENCLAW, HERMES, PI, AND OPENHUMAN. EVERY OTHER MCP CLIENT GETS IT FOR FREE. agentmemory connect <agent> AUTO-WIRES THEM ALL.
FIRST-CLASS PLUGINClaude CodeFROM Anthropic
12 hooks + MCP + skills
NATIVE PLUGINCopilot CLIFROM GitHub
11 hooks + MCP · framed stdio
NATIVE PLUGINCodex CLIFROM OpenAI
6 hooks + MCP · native plugin
GATEWAY PLUGINOpenClawFROM openclaw
onSessionStart · onPreLlmCall · onPostToolUse · onSessionEnd
FIRST-PARTY INTEGRATIONHermesFROM Nous Research
Python plugin · yaml config
Native plugin + MCP
NATIVE BACKENDOpenHumanFROM tinyhumansai
Native Memory trait backend (Rust)
ONE INSTALL. #
ANY AGENT.
RUNS ON YOUR MACHINE. DATA STAYS LOCAL. BRING YOUR CLAUDE SUBSCRIPTION — OR POINT IT AT ANTHROPIC, GEMINI, MINIMAX, OR OPENROUTER.
ONE MCP JSON FITS ALMOST EVERYTHING. PICK YOUR AGENT ON THE LEFT, OR PASTE THE UNIVERSAL CONFIG ON THE RIGHT.
CURSOR / VS CODE ARE ONE-CLICK VIA DEEPLINK. OTHERS COPY THE RIGHT SNIPPET DIRECTLY TO YOUR CLIPBOARD.
{
"mcpServers": {
"agentmemory": {
"command": "npx",
"args": ["-y", "@agentmemory/mcp"],
"env": {
"AGENTMEMORY_URL": "http://localhost:3111"
}
}
}
}