// Agent profile
Engram (Gentleman-Programming/engram) is a persistent memory system for AI coding agents, shipped as a single Go binary with zero dependencies. It uses SQLite with FTS5 full-text search for storage, exposing 20 MCP tools for save, search, session management, and conflict detection. Agents call mem_save with structured What/Why/Where/Learned data, which Engram persists and indexes. Next session, the agent searches memory and gets relevant context automatically. Agent-agnostic — works with Claude Code, Gemini CLI, Codex, Cursor, Windsurf, OpenCode, and any MCP-compatible tool. Includes an interactive TUI with Catppuccin Mocha theme, git sync for cross-machine memory sharing, and optional cloud replication. 5.3K GitHub stars, MIT-licensed.
Engram is a persistent memory system that gives AI coding agents recall across sessions. Named after the neuroscience term for a physical memory trace, it provides structured storage that agents can write to and query from automatically. The core architecture is a single Go binary backed by SQLite with FTS5 indexing — no external dependencies, no cloud requirement, no container setup.
Structured memory format: Agents save memories with a What/Why/Where/Learned structure, which Engram indexes for both keyword and semantic retrieval. The 20 built-in MCP tools cover the full lifecycle — saving, searching, tagging, session management, and conflict detection when multiple agents write contradictory information.
Agent-agnostic design: Engram works with any tool that supports MCP via stdio transport. Claude Code, Gemini CLI, Codex, Cursor, Windsurf, OpenCode, and Antigravity are all supported out of the box. Adding support for a new agent requires only pointing it at Engram’s MCP endpoint.
Cross-machine sync: Memories can be synced across machines via git, treating the SQLite database as a versioned artifact. For teams, optional cloud replication keeps multiple developers’ agent memories in sync.
Interactive TUI: A terminal-based interface lets developers browse, search, and manage memories directly. The TUI uses the Catppuccin Mocha color scheme and supports filtering by type, date, and agent source.
Developers use Engram to maintain project context across coding sessions — architectural decisions, debugging findings, API quirks, and codebase conventions persist even when the agent’s context window resets. Teams use git sync to share institutional knowledge between agents running on different machines. The conflict detection feature is particularly useful in multi-agent setups where two agents might record contradictory information about the same topic.
Engram requires Go for installation from source, though pre-built binaries are available. The FTS5 search is keyword-based by default; semantic search requires an LLM backend. Memory quality depends on what the agent writes — garbage in, garbage out.
Developers who run AI coding agents across multiple sessions and want their agent to remember what it learned yesterday. Teams that want shared agent memory without a cloud service. Anyone tired of re-explaining project context at the start of every session.
Builder.io's open-source framework for building agent-native applications — shared actions, SQL-backed state, identity, tools, skills, jobs, observability, and UI surfaces that all work together.
CLI orchestration tool that installs and configures platform connectors for 16+ social and web sources — Twitter/X, Reddit, YouTube, GitHub, Bilibili, and more. Zero API fees.
Persistent memory layer for AI coding agents — benchmark-backed (95.2% on LongMemEval-S), 92% fewer tokens per session vs full-context pasting, zero manual memory.add() calls.