Skip to content
Stork.AIstork.ai
AI Tool

Agentmemory Review

Agentmemory is an open-source persistent memory layer designed for AI coding agents, enabling them to retain context and learn across sessions.

shipped May 18, 2026aifreemium
ai
Agentmemory - AI tool for agentmemory. Professional illustration showing core functionality and features.

Why it matters

1Achieves 95.2% retrieval R@5 accuracy.
2Reduces token usage by 92% compared to traditional methods.
3Operates with zero external databases, supporting local-first deployment.
4Gained significant traction, hitting #1 on GitHub trending with 1,048 stars in 24 hours in May 2026.

Stork’s verdict on Agentmemory

Agentmemory provides local-first persistent memory for agents, but requires developer integration via its API.

Specs

API Docs

View Documentation →

API Available

Yes, public API

overview

What is Agentmemory?

Agentmemory is an open-source persistent memory layer tool developed by Agentmemory that enables AI coding agent developers to provide persistent memory for AI coding agents. It silently captures agent actions, compresses them into searchable memory, and injects relevant context into future sessions. This tool addresses the 'goldfish-memory problem' prevalent in AI agents, where context is typically lost after a session concludes. Agentmemory aims to establish a unified memory store accessible across diverse AI coding agents, enhancing their ability to learn and adapt over time. Its core functionality involves capturing agent interactions, compressing them into a searchable format, and leveraging this memory to inform subsequent sessions, thereby eliminating the need for repetitive explanations or context re-establishment.

features

Key Features of Agentmemory

Agentmemory provides a robust set of features designed to enhance the capabilities of AI coding agents by offering persistent, efficient, and context-aware memory management. Its architecture prioritizes local operation and high retrieval accuracy, ensuring agents can access relevant information rapidly and effectively.

  • Persistent memory for AI coding agents, enabling context retention across sessions.
  • Achieves 95.2% retrieval R@5 accuracy for relevant context.
  • Reduces token consumption by 92% through efficient memory compression and retrieval.
  • Operates with zero external databases, facilitating local-first installation and deployment.
  • Ensures compatibility with every agent, offering broad integration potential.
  • Provides a comprehensive API for programmatic access and integration (API Docs URL: https://agent-memory.dev/#rest-endpoints).
  • Supports local execution, allowing agents to run anywhere without cloud dependencies.
  • Automatically captures every agent session, building a continuous memory stream.
  • Recalls relevant information in milliseconds, minimizing latency for agents.
  • Injects relevant context into future sessions, improving agent performance and reducing redundant inputs.
  • Includes an OpenAI-compatible LLM provider, supporting endpoints like OpenAI, Azure OpenAI, DeepSeek, and Ollama.
  • Utilizes a retrieval pipeline combining BM25, vectors, and a knowledge graph fused with reciprocal rank fusion for efficient context retrieval.

use cases

Who Should Use Agentmemory?

Agentmemory is primarily designed for developers and teams working with AI coding agents who require robust, persistent memory solutions to overcome the limitations of stateless AI models. Its capabilities are particularly beneficial in scenarios demanding continuous learning, context maintenance, and collaborative agent workflows.

  • AI coding agent developers: For providing persistent memory and maintaining context across agent conversations and sessions, enabling agents to remember project architecture and past decisions.
  • Developers building AI agents: To enable agents to learn from past interactions and experiences, supporting the execution of complex, multi-step tasks by agents.
  • Teams utilizing multiple AI agents (e.g., Claude Code, Cursor, Codex CLI): To facilitate shared memory across various agents, people, and tools, fostering collaborative development environments.
  • Developers seeking to eliminate repetitive explanations for AI coding assistants: By allowing agents to recall previously encountered bugs and solutions, reducing redundant input.
  • Developers requiring efficient context retrieval: Leveraging its retrieval pipeline combining BM25, vectors, and a knowledge graph for precise and rapid information access.

pricing

Agentmemory Pricing & Plans

Agentmemory operates on a freemium model, making its core functionalities accessible without upfront costs. While in beta, default capacity and rate limits apply to memory stores. Users requiring higher limits or specific configurations are advised to contact support. The system also supports configurable rate limiting, including token-based and request-based limits, which can be applied per key or per time window, offering flexibility for various deployment scales.

  • Freemium: Free (includes default capacity and rate limits in beta; configurable rate limiting available upon request).

Similar Tools

Agentmemory vs Competitors

Agentmemory distinguishes itself in the AI agent memory landscape through its open-source nature, local-first approach, and emphasis on benchmarked retrieval accuracy and token efficiency. It competes with several established and emerging solutions, each offering distinct architectural and feature sets.

1
Mem0On StorkCompare

Mem0 provides a dedicated, intelligent memory layer for AI applications with multi-level memory scopes and hybrid memory retrieval.

Similar to Agentmemory, Mem0 focuses on enhancing AI agents with personalized, persistent memory, offering a fully managed service option and SDKs. It explicitly supports multi-level memory (user, session, agent) and a graph layer for relationships, which expands on Agentmemory's core retrieval and token efficiency.

2
ZepOn StorkCompare

Zep is a long-term memory store designed specifically for conversational AI, excelling in extracting facts, summarizing conversations, and providing temporal and semantic search.

Zep primarily targets conversational AI applications, emphasizing temporal relationships and progressive summarization, which offers a more specialized focus compared to Agentmemory's broader application for coding agents. It provides both semantic and temporal search capabilities.

3
Letta (formerly MemGPT)↗

Letta employs an operating system-like architecture to manage a 'virtual context,' allowing agents to access significantly more memory than typical context window limits.

Letta's approach to memory involves managing a 'virtual context' and providing explicit control over memory blocks, representing a different architectural paradigm than Agentmemory's focus on retrieval efficiency. It is open-source and self-hosted, aligning with Agentmemory's '0 external databases' for potential self-hosting.

4
Supermemory.ai↗

Supermemory.ai offers a comprehensive five-layer memory solution, including user profiles, a memory graph, and a custom vector graph engine for deep understanding and context.

Supermemory.ai positions itself as an all-in-one memory solution with multiple integrated layers, aiming to replace several services. This contrasts with Agentmemory's focus on a single, efficient memory layer without external databases, suggesting Supermemory.ai might offer a broader, more complex suite of memory functionalities.

Visit Agentmemory↗
Connect
𝕏
X / Twittertwitter.com/agentmemory

AI Reputation Report

Is Agentmemory yours?

ChatGPT, Perplexity, Gemini, Claude & Grok answer buyer questions about Agentmemory every day. See whether they name Agentmemory — or send buyers to a rival.

See what AI saysfree preview