agentmemory review
A self-hosted memory runtime that captures agent sessions and recalls context across tools.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
agentmemory is an open-source, local-first memory runtime for developers who want coding agents to retain context across sessions. It captures prompts, tool calls, results and session events, then stores durable typed memories, lessons and session summaries. The runtime and browser viewer run on the developer’s machine, and multiple agents can share the same memory store. That makes it a fit for solo developers and small teams working across coding-agent tools, especially those who prefer local deployment over an account-based hosted service.
Its core is session capture and retrieval rather than a single chat history. Hybrid search combines BM25 keyword retrieval with vector semantic retrieval, while knowledge graph extraction and graph browsing offer another way to explore relationships among stored information. A browser viewer includes live stream, replay and health status; the runtime also provides a REST API and MCP server. Offline spool and crash-safe recovery address interruptions, while secret redaction and loopback authentication are enabled by default. Connectors include Claude Code, Codex CLI, Cursor, OpenCode, Devin, OpenClaw, Hermes, Gemini CLI, Claude Desktop and Aider.
The runtime is Apache-2.0 open source and has a free plan; according to the vendor’s notes, there is no hosted tier. That tradeoff puts deployment and operation on the developer, so teams seeking a managed service should look elsewhere. The breadth of capture, retrieval, replay and agent hooks is useful when the goal is persistent coding-session context, but it may be more infrastructure than someone needs for a narrow, single-agent workflow. Documentation is the listed support channel. Choose agentmemory if local control and shared memory across coding agents matter; choose a hosted option if avoiding self-hosting is the priority.
agentmemory pros and cons
- Where it wins
- Captures prompts, tool calls, results and session events via hooks
- Combines BM25 keyword and vector semantic retrieval with graph browsing
- Local REST/MCP server supports multiple coding-agent integrations
- Where it doesn't
- Requires a local, self-hosted runtime rather than a hosted service
- Its focus is coding-agent memory, not general-purpose agent infrastructure
- Support channels listed are documentation only
agentmemory fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. Vendors can pay for top positions in our rankings and for a place on other products' pages, and we may earn a commission when you click some links. How we rank.
Last updated · How we research and update