AI Agent Tools: Memory & Context

graphify146.4k★

Graphify turns a codebase, plus its docs, SQL schemas, PDFs and images, into a knowledge graph your AI coding assistant queries instead of grepping through files. Code is parsed locally with tree-sitter: no LLM, no API key, nothing leaves your machine. The CLI is Apache 2.0 and free. Install with uv tool install graphifyy (double y; other graphify packages on PyPI aren't theirs), then graphify install adds a /graphify skill to Claude Code, Cursor, Codex and 20+ others. Docs, PDFs and images go through a model, so a docs-heavy repo costs tokens. Code alone costs nothing. Graphify Labs now sells Graphify Cloud: one always-on graph across repos, with PR review and formal verification. Free covers 3 repos, Pro is $10/month billed yearly, Teams starts at $20/seat billed yearly. Solo and small teams: the free CLI. Many repos: that's the Cloud pitch. The catch is the moving target. It changed orgs, relicensed from MIT to Apache 2.0 in July, and is now a YC company shipping almost daily. The "80% fewer tokens" headline is about the paid Cloud, not the free CLI.

OpenViking42.0k★

OpenViking is a context database built specifically for AI agents. Instead of the usual pile of vector embeddings, it organizes an agent's memory, retrieved documents, and skills the way a filesystem organizes files: directories you can navigate, tag, and search recursively. It is open source under AGPL-3.0. Setup is heavier than a typical pip install. You will need Python 3.10+, a Rust toolchain for building the RAGFS component, and a C/C++ compiler for the core extensions, on top of wiring up a vision-language model and an embedding model (Volcengine Doubao, OpenAI, or others) before it does anything useful. Once running, it compresses conversation history automatically and shows you the retrieval path it took to answer a query, so debugging why an agent recalled the wrong thing is actually possible. There is a hosted demo, OpenViking Studio, if you want to try the retrieval and memory features before committing to the build. Solo devs experimenting with agent memory: worth the setup time if you are hitting the limits of flat vector search. Teams building production agents: the tiered loading, it only pulls in the context tier it actually needs, is the real selling point for cutting token costs at scale. It is AGPL-3.0, not MIT or Apache. If you build a hosted product on top of OpenViking and offer it as a network service, the license requires you to open source your modifications too. That is a real constraint for anyone planning to build a commercial product on this, not just an internal tool.

agentmemory30.3k★

agentmemory gives your coding agent a memory that survives the session. It hooks into Claude Code, Cursor, Codex, Gemini CLI and more, records what the agent did, and hands the relevant pieces back next time, so you stop re-explaining your stack. It's Apache-2.0, runs on your machine, and it's free. Setup is one npx -y @agentmemory/agentmemory@latest, which wires your agents and starts a local server plus a pinned iii engine binary. No external database, but it wants Node 20 and four local ports, and native Windows means installing the engine by hand. Default search is keyword-only; EMBEDDING_PROVIDER=local adds free on-device semantic recall. There is no paid tier. Your only bills are an optional LLM provider and a small server if a team wants shared memory. Solo: free, an easy win. Small teams: free. Larger teams: free, but you're self-supporting a pre-1.0 project. The catch: the "#1" in the tagline is marketing. The headline 95.2% is retrieval recall on LongMemEval-S, not that benchmark's official QA score, and the project's own report lists MemPalace higher at 96.6%. Compare Mem0, Zep and Letta yourself. Also, the required iii engine is Elastic License 2.0, not Apache.

get-shit-done-redux11.5k★

GSD Core keeps AI coding assistants reliable over long sessions. The problem it targets is context rot, the quality drop that happens as an agent fills its context window and starts forgetting what it was doing. It runs the heavy research, planning and execution in fresh subagents so your main session stays lean. It's the community continuation of the original Get Shit Done project, MIT licensed and free. Install is one command: npx @opengsd/gsd-core@latest. It works with Claude Code, OpenCode, Codex, Copilot, Cursor, Windsurf, Antigravity CLI, Kimi CLI and others. The workflow is a new-project command plus a five-step loop: discuss, plan, execute in parallel waves, verify, ship. Releases land every week or two, with a migration path off the deprecated original package. Solo developers who run long agent sessions and watch quality degrade get the most out of this. Small teams adopting a shared agent workflow can standardize on the same loop. It's free at every scale; the cost is learning the discipline it imposes, which is the point. The catch: it's methodology wrapped in tooling and only helps if you commit to the loop. Bolt it onto a chaotic workflow and you get chaos with extra steps. A hosted gsd-cloud is announced as coming soon with no pricing yet, so watch whether the free core stays the whole product.

fast-jev-compaction8.0k★

fast-jev-compaction changes what happens when a long Claude Code session runs out of room. The built-in behavior summarizes old conversation, which loses detail. This plugin asks Jev, a decision model from TypeSafe AI, which old tool calls and outputs are stale, deletes only those, and keeps every user and assistant message word for word. MIT licensed. Install goes through the Claude Code plugin marketplace, straight from the repo. It needs an environment flag for function hooks, which the README calls an early-access Claude Code feature, plus a TypeSafe API key. Jev bills $0.042 per million input tokens and is waitlisted. When Jev errors out or does not cut enough, the plugin falls back to the normal summary. Solo developers living in long Claude Code sessions are the audience, once they have Jev access. Teams should think harder, because the plugin sends your session transcript to TypeSafe for scoring. The catch: do not run the npm install line in the README. As of September 21, 2026, the fast-jev-compaction name on npm belongs to a different account publishing its own fork with extra command line tools. Nothing in it looked malicious on a static read, but it is not the author's package. The marketplace path pulls from the repo and is unaffected. Token sizes are also estimated from character counts, and users report the hooks not loading on some Claude Code versions.

engram7.9k★

Engram is a memory layer for AI coding agents. Your agent forgets everything when a session ends; Engram stores the bug fixes, decisions and conventions it learns in a local SQLite file with full-text search, then hands them back next session. It works with Claude Code, Codex, Cursor and any other agent that speaks MCP (the plug-in standard agents use to call outside tools). MIT licensed and free. Setup is easy: brew install gentleman-programming/tap/engram, then engram setup claude-code (or your agent). One Go binary, no Node, Python or Docker. Sharing across machines works through Git sync, or Engram Cloud, which is a server you host yourself on Docker and Postgres, not a paid service. There's no paid tier anywhere. Solo: free, five minutes. Small teams: free with Git sync or one self-hosted Cloud box. Larger teams: free, but someone owns the server, tokens and upgrades. The catch: memory is only as good as what the agent chooses to save, and junk memories pollute later sessions. "Personal" memories are a search filter, not a privacy wall: enroll a project for sync and they go to everyone with access. Releases land almost daily, so pin versions. claude-mem is the closest alternative.

OptMem2.4k★

OptMem gives your coding agent long-term memory using a short prompt and one Python script. Memories go into an append-only log of fixed-width records, so every lookup is a single disk seek, with no database or vector store. The agent summarizes older memories into a binary tree as it goes and searches the full log by regex when it needs detail. It's from Victor Taelin, of HVM and Bend. Install is a curl-to-shell script that drops one dependency-free Python file into ~/.optmem and runs init; neither the script nor the tool makes network calls. Commands cover waking up, noting, napping (answering pending merges), recall, zoom and forget. The summaries are written by your agent, so it spends its own tokens on compression. Free to download, but not open source: there's no license file, so no reuse or redistribution rights are granted, and a license question in the issues has gone unanswered since August. Solo devs experimenting with agent memory can try it. Teams shouldn't build on code they have no license to. The catch: the prompt tells the agent to record anything it learns about your life, even indirectly, into a plaintext log it's told never to edit or delete. Development stopped at the end of July, and the install script pulls from an unpinned main branch.

vibecode-pro-max-kit1.2k★

vibecode-pro-max-kit is an installable bundle of agents, skills, and lifecycle hooks for Claude Code, Cursor, Codex, and similar AI coding environments. The pitch is a spec-driven workflow that pushes you through research, innovate, plan, and execute phases with explicit approval gates instead of letting the agent run free. MIT-licensed, free, installed by curl. Setup is a 30-second curl that drops 12 specialized agents, 31 skills, and hooks into your project's .claude/ directory. The framework's bet is that disciplined phase gates beat raw agent autonomy: you don't get past planning without approving the plan, and you don't get past execution without a spec to check against. Works across any tech stack since it's harness-level, not language-level. For solo developers who keep losing context with vibe-coded agent runs, this is a structured replacement. Small teams adopting it together get a shared coding workflow without having to enforce one manually. Larger orgs will want to read what the agents actually do before standardizing on someone else's prompt library, but the core idea (gate the agent, don't free-run it) is good practice. The catch: this is an opinionated workflow, not a tool. Skip the gates and you've installed 12 agents producing noise. The upstream also changes whenever the maintainer changes their mind, which is its own kind of dependency. If you want to enforce spec-driven coding with AI, this is one way; if you want the agent to be fast and loose, skip it.

recall769★

Recall fixes the annoying part of working with Claude Code across sessions: re-explaining your project every time you come back. It captures what happened in a session, the prompts, the responses, the files you touched, the commands you ran, and boils it down to a short summary the agent can read next time to pick up where you left off. MIT licensed, free, and it runs entirely on your machine. The clever bit is that it summarizes locally with classic text-ranking algorithms, TF-IDF and TextRank, so there is no extra LLM call and no token cost for the summary itself. It drops two files in a .recall/ folder: an append-only history log and an auto-generated context digest with your goal, key files, and next steps. It even redacts common secret patterns before writing anything to disk. No setup, works offline. This is a Claude Code plugin, not an MCP server, so it slots into the workflow without extra infrastructure. Solo developers juggling several projects get the most out of it: a 1 to 2K token digest is far cheaper than pasting your whole project history back in, and it works with whatever Claude subscription you already have. The catch is that a TF-IDF summary is compression, not understanding. It surfaces what looks important by word frequency, which is usually enough to orient the agent but occasionally misses the one detail that mattered. Skim the context file before you trust it to have remembered everything.