Engrim
NexusAi Summary: Engrim is a local-first episodic memory engine for coding agents, storing high-signal decisions in SQLite and reloading ~4k characters per turn. It reduces context bloat, works offline, and syncs memory across Antigravity, Claude Code, Cursor, Codex CLI, OpenCode, and Windsurf.
NexusAi Overview
On each invocation, Engrim injects a small, priority‑ordered memory pack that captures active decisions, constraints, and a clear resume pointer. As agents work, new facts and choices are written to the store through lifecycle hooks or CLI tools. Retrieval blends lexical and vector signals for precise, low‑latency recall without cloud dependencies.
How Engrim Works
Engrim suits principal engineers, staff developers, agent toolsmiths, and founders building agentic coding workflows. It shines in long refactors, rule‑heavy systems, test stabilization, and handoffs across tools where resets are common. Teams demanding offline operation and strong privacy—such as finance, healthcare, and regulated industries—benefit from sub‑1% context footprints, deterministic behavior across hosts, and the ability to audit what the agent actually remembered and why.
- Injects a compact, priority‑ordered memory pack to restore context after clears.
- Shares the same SQLite store across multiple coding agents and environments.
- Retrieves memory using FTS5 bm25 plus static model2vec embeddings on CPU.
- Tracks provenance per record to audit origin across tools and sessions.
- Offers self‑healing diagnostics to validate hooks, embeddings, and database integrity.
Why It Stands Out
Editor’s Take
Install via pip, then run a one‑command setup to auto‑detect and wire supported environments. Projects immediately benefit from a budgeted memory pack at boot, and agents can add or recall records through hooks, MCP tools, or CLI commands. Use the review flow before clearing to ensure important decisions are captured, then freely reset sessions—Engrim restores focus on the next turn. The diagnostic tool validates WAL integrity, embedding coverage, and hook health, and portable fallbacks keep workflows resilient across machines. Backups and merges let you move or federate stores safely.
NexusAi: By turning curated memory into a portable, local substrate, Engrim restores agent focus, slashes token waste, and enables true cross‑tool continuity without compromising privacy.
Getting Started
Engrim is a focused memory substrate for engineering agents: small, precise, and portable. It minimizes context overhead, preserves architectural intent, and enables continue‑as‑clear workflows across tools. It is not a general knowledge base; it is a durable, audited working memory for code projects. Paired with disciplined capture, Engrim delivers sharper reasoning, lower costs, and reliable multi‑agent continuity.
Open the tool and review its core product experience.
Create your account or access your existing workspace.
Use your own task to judge speed, quality, and fit.
Check similar AI tools before making a final decision.
Engrim
Engrim is led by Tim Gordon, a practitioner building agentic tooling from real production constraints. The project focuses on durable, local memory for complex software work, emphasizing offline operation, disciplined retrieval, and cross‑tool continuity over cloud services and transcript replay.
Engrim is a universal, cross‑model episodic memory layer purpose‑built for software engineering agents. Rather than replaying long transcripts, it persists curated decisions, constraints, and state in a single local SQLite store, then injects a compact ~4k‑character memory pack on every turn. Hybrid retrieval pairs FTS5 bm25 with static model2vec embeddings for fast, offline recall on CPU. Hooks and an MCP server let Antigravity, Claude Code, Cursor, Codex CLI, OpenCode, and Windsurf share identical project memory, with origin provenance and a resume pointer for seamless restarts after clearing context. A self‑healing diagnostic tool verifies WAL integrity, embedding coverage, and hook wiring to prevent silent failures. With private, local storage and zero telemetry, Engrim sharpens reasoning, slashes per‑turn context costs, and eliminates vendor lock‑in for agentic coding workflows.
Visit Engrim
Comments (0)
No Comments Found