Skip to content
Sign in
  • Browse all toolsFilter by category and pricing
  • Browse categoriesWriting, code, image, audio and more
  • Compare toolsPut up to three side by side
  • Read the blogGuides and comparisons

agentmemory

Persistent memory for AI coding agents, with shared session history, searchable fixes and a local viewer.

agentmemory is a persistent memory project that captures coding sessions and retrieves decisions, fixes and project history across supported agents. Designed for developers using AI coding tools, agentmemory is available under the Apache-2.0 license, with account-free keyword recall and optional free local embeddings for semantic search.

Updated 9 Nov 20266 sections

At a glance

Pricing
FreeApache-2.0 open-source project. Keyword recall needs no account or model API key, and local semantic recall is available free on-device.
Added
9 Nov 2026
agentmemory website screenshot

What is agentmemory?

agentmemory is an open-source persistent memory project for AI coding agents. agentmemory captures supported coding activity and turns saved decisions, fixes and lessons into searchable project history. Developers can use the same memory store across supported tools rather than treating each coding session as an isolated conversation.

agentmemory extends the LLM Wiki pattern described on the repository page with confidence scoring, lifecycle management, knowledge graphs and hybrid search. The project includes versioned memories, lessons, actions and import/export capabilities. A local viewer makes saved information and original sources available for inspection.

What agentmemory can do

agentmemory combines session capture, retrieval and memory management for existing coding workflows. Native integrations record supported events, while search helps agents recover earlier work. Retrieval can use keyword matching without a model API, optional embeddings for semantic matching, and structural graph matches when the required graph data already exists.

  • agentmemory records supported session events through native hooks and plugins.
  • agentmemory scrubs secrets and supports offline replay during capture.
  • agentmemory retrieves saved memories through BM25 keyword search.
  • agentmemory supports local or remote embeddings for semantic recall.
  • agentmemory retrieves project history, lessons and session handoffs.
  • agentmemory lets users browse, edit, export and forget memories in the live viewer.

How agentmemory works

agentmemory starts through an npx-based setup that asks users to select agents and either a provider or keyless mode. Setup writes configuration, starts the memory server and offers a global command installation. The API and live viewer bind to loopback by default, and memories stay outside the repository by default.

agentmemory connects tools through native plugins, hooks, MCP or a REST API, depending on the integration. Supported examples include Claude Code, Cursor, Codex CLI, GitHub Copilot CLI, Gemini CLI and Aider. Supported agents share the same memory server, and additional agents can be connected with the agentmemory connect command.

Who agentmemory is for

agentmemory serves developers who want coding agents to retain useful context across sessions and tools. The documented workflow focuses on recovering project decisions, fixes and lessons while keeping saved memory inspectable. Shared memory also supports continuing work with another agent rather than reconstructing the project's history from scratch.

  • agentmemory helps developers recall fixes from previous coding sessions.
  • agentmemory supports project-history sharing across supported coding tools.
  • agentmemory enables session handoffs between agents.
  • agentmemory supports reviewing and editing saved context through the local viewer.

agentmemory pricing and plans

agentmemory is distributed as an Apache-2.0 open-source project through its repository and npm package. Keyword recall requires neither an account nor a model API key. Free on-device semantic recall is available by enabling the local embedding provider; the initial embedding request downloads a model before subsequent inference runs locally.

agentmemory also supports remote embeddings and opt-in model-written observation compression with a provider. The page does not list subscription tiers or provider prices. Local installation requires Node.js 20 or newer with npm and npx, with additional setup requirements documented for macOS, Linux and Windows.

agentmemory strengths and limitations

agentmemory combines shared cross-agent memory, keyless keyword retrieval and direct inspection of saved information. The repository distinguishes its retrieval component benchmark from full-system answer quality and competitor rankings. This distinction matters when assessing the published results: retrieval from supplied sessions does not establish overall answer quality or a head-to-head ranking.

  • agentmemory disables vectors in default keyless mode, so semantic queries can return no results.
  • agentmemory requires an initial model download before local semantic recall is available.
  • agentmemory enables model-written observation compression only with explicit configuration and a provider.
  • agentmemory captures supported session events according to the available agent integration.

agentmemory FAQ

What is agentmemory?

agentmemory is an open-source persistent memory project for AI coding agents. agentmemory captures supported coding sessions, makes decisions and fixes searchable, and lets agents retrieve project history later. Supported tools can share a memory server, while a local viewer exposes saved memories for inspection and editing.

Is agentmemory free?

agentmemory is available as an Apache-2.0 open-source project, with keyword recall that needs no account or model API key. agentmemory also offers free on-device semantic recall through local embeddings. Remote embeddings and provider-based observation compression are optional, and the page does not specify provider prices.

Which AI coding agents does agentmemory support?

agentmemory supports Claude Code, Cursor, Codex CLI, GitHub Copilot CLI, Gemini CLI, OpenCode and other listed tools. agentmemory connects through native plugins, hooks, MCP or REST API access, depending on the integration. Agents that speak MCP or HTTP can use the shared memory server.

Can agentmemory work without an API key?

agentmemory supports keyword recall without a model API key through BM25 search. Default keyless mode disables vector embeddings, so semantic queries may return no results. agentmemory can add free on-device semantic recall by enabling the local embedding provider and allowing the initial model download.

How do you install agentmemory?

agentmemory installs through the documented npx setup using Node.js 20 or newer with npm and npx. Setup asks users to choose agents and a provider or keyless mode, writes configuration and starts the memory server. agentmemory also documents a demo for testing recall and a command for connecting additional agents.

Where does agentmemory store memories?

agentmemory stores memories outside the repository by default, using platform-specific data directories on macOS, Linux and Windows. agentmemory supports a custom data directory and separate instances. The documentation instructs users to retain the same data directory when restarting or upgrading.

From the community

No reviews yet

Be the first to review this tool. What you write helps the next person choose.

More tools like agentmemory

deja-vu

www.github.com

Local memory and session search for AI coding agents, with MCP recall, hooks and repository-shared decisions.

Free

Claude-Mem

www.github.com

Persistent memory for coding agents that captures session activity and retrieves relevant context in future sessions.

Freemium

claude-supermemory

www.github.com

Persistent memory for Claude Code with automatic recall, conversation capture, and shared repository context.

Freemium

MemPalace

www.github.com

Local-first AI memory with verbatim conversation storage, semantic search, CLI access and MCP integration.

Free