Summary

Engram provides persistent memory for AI coding agents, solving session-based amnesia. It offers an agent-agnostic, single Go binary with zero runtime dependencies, storing memories in SQLite via FTS5. Accessible via CLI, HTTP API, and TUI, Engram enables AI agents to retain and recall context across sessions, enhancing productivity.

How It Works

Engram relies on agent-curated memory. Agents proactively save structured summaries of significant work (e.g., bug fixes, decisions) using the mem_save MCP tool, avoiding raw tool call capture for cleaner data. Engram persists these to SQLite with FTS5 indexing. Retrieval uses progressive disclosure: search yields compact results, followed by timeline context, then full content.

Quick Start & Requirements

Installation is recommended via Homebrew (brew install gentleman-programming/tap/engram). Alternatively, download binaries or build from source (go install ./cmd/engram). Agent integration uses commands like engram setup opencode. Building from source requires Go 1.25+. Engram has zero runtime dependencies, including SQLite via pure Go.

Highlighted Details

• Agent-Agnostic: Works with any MCP-compatible agent (OpenCode, Claude Code, Cursor, etc.).
• Minimalist Architecture: Single Go binary with SQLite/FTS5; no Node.js, Python, Docker, or vector databases needed.
• Progressive Disclosure: Token-efficient retrieval (search -> timeline -> full content).
• Memory Hygiene: Topic upserts, soft-deletes, and exact deduplication maintain data integrity.
• Git Sync: Share memories across machines via compressed chunks and a manifest, avoiding merge conflicts.
• Interactive TUI: A terminal UI for browsing, searching, and navigating memories.
• claude-mem Alternative: Simpler, agent-agnostic approach focusing on agent-curated summaries.

Maintenance & Community

The provided README does not detail notable contributors, sponsorships, or community channels.

Licensing & Compatibility

Released under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

Engram's effectiveness depends on agents proactively saving meaningful information. Resilience against context resets (compaction) requires correctly configured "Surviving Compaction" instructions; otherwise, memory persistence during severe context loss is not guaranteed.