Mengram

About Mengram

Mengram is an AI memory API that stores three types of memory: semantic (facts), episodic (events), and procedural (workflows). It extracts all three types in a single call and can persist the sequence of steps an agent took so future runs can reuse successful workflows.

Review

Mengram provides a focused solution for long-term agent memory by combining facts, dated events, and replayable procedures into one service. The project is open-source under the Apache 2.0 license and emphasizes a quick setup experience for developers.

Key Features

• Three memory modalities (semantic, episodic, procedural) with a single extraction API.
• Automatic capture of completed task steps as reusable procedures, including success/failure tracking.
• Built-in conflict detection and staleness management, plus date-range filtering for episodic queries.
• Drop-in replacement for typical memory and retrieval layers and integrations with common agent frameworks and LLM toolchains.

Pricing and Value

Mengram is free and open-source under the Apache 2.0 license, making it accessible for prototyping, research, and production use when teams prefer self-hosting or code-level customization. Its value proposition lies in reducing implementation effort by handling contradictions, decaying stale items, and storing procedural workflows out of the box, which can shorten development time for agent-driven applications.

Pros

• Comprehensive support for three distinct memory types in one API.
• Procedural memory allows agents to reuse previously successful step sequences.
• Automatic mechanisms for conflict detection and memory staleness reduce maintenance burden.
• Single-call search across all memory types simplifies retrieval logic.
• Open-source with a permissive license encourages customization and inspection.

Cons

• New project with a smaller community and fewer third-party integrations compared with established alternatives.
• Integration with existing LLM and embedding services may require additional configuration depending on your stack.
• Organizations needing dedicated enterprise support or managed SLAs may find the current offering limited.

Mengram is a strong fit for developers building AI agents that need structured, persistent memory and for teams that prefer an open-source solution they can extend. It is especially useful for prototypes and internal tools that benefit from procedural replay; teams seeking a mature, fully managed enterprise product should evaluate support and operational needs before committing.