Cadenya
Paid
Developers building agentic workflows
Brain Memory
FreeOpen SourceSelf-Hosted
Every time you switch models or start a new session, your agent forgets what it decided last week — and you spend the first twenty minutes re-explaining context it already earned. Brain Memory is a file-system-backed memory layer that makes those decisions persist, decay realistically, and recall deterministically across every agent and CLI you point at it.
Similar tools
JevProxyFreemiumThe proxy sits in front of your Cursor or Claude Code agent fleet and classifies incoming tool…
CadenyaPaidCadenya is an agent runtime built around a layered model: tools connect once through MCP…
WorldsFreemiumWorlds spins up a sealed twin of the external systems your agent touches — Stripe and Zendesk…
EmemFreemiumemem stores facts as short, signed tokens — each one a content-addressed handle that any agent…
Get This Tool
License: MIT Any use incl. commercial
Local-run terms: Install via npm, writes to ~/.brain/ directory, full source available under MIT for commercial and private use.
Pricing
- Model
- Free
Summary
Brain Memory stores agent decisions as Markdown files with YAML frontmatter, organized in a directory tree you can browse in any file explorer — no opaque vector database, no embeddings you cannot audit. Strength decays on an Ebbinghaus exponential curve and rebuilds each time a memory is recalled, so the architecture you revisited three times stays sharp while the one-off experiment fades. The benchmark vendor cites shows 100% recall on a 1,000-distractor haystack where BM25 and vector retrievers both score zero — a meaningful gap for long-running coding projects. The tool is at v0.1.0, MIT-licensed, and ships as an npm global install. That version number is not a warning to ignore: the sleep consolidation pipeline and the cross-agent sync model are genuinely novel, which means the surface area for early-stage bugs is wider than a mature retrieval library.
Bottom line: Bet on this for a multi-session coding project where three different CLI agents need to share one memory store without re-prompting — but expect to hit rough edges if you need a stable API contract or an integration outside the named agent list.
Find your fit
Before you choose Brain Memory
More options in Agent Frameworks. Compare pricing, platforms, and features side by side.
Agent-Talk
Free · Open source
Cross-model agent collaboration
Pros
- Deterministic recall scoring across every supported agent and model, so switching from Claude Code to Codex CLI mid-project returns identical memory results — no per-agent re-tuning required.
- Decay rates differ by cognitive type (episodic vs. procedural), which means architecture decisions that matter accumulate strength while one-off debugging notes fade on their own — without manual curation.
- Plain Markdown files in a directory tree, so memory is auditable in any file explorer and syncs via git, Dropbox, or iCloud without a proprietary backend — which means no vendor lock-in and no account required.
- Token-efficient recall loads only the memories relevant to the current query rather than dumping the full store, with the vendor citing a 31% token reduction versus full-skill loading — so context windows stay usable on long projects.
- Contradiction handling via decay-weighted recency means when a preference flips (tabs, then spaces, then tabs again), the latest version wins automatically rather than requiring a manual memory edit.
Cons
- No public API is available, so any system outside the named agent list — a custom Python agent, a backend service, an internal tool — has no supported path to read or write memory; teams that need programmatic access build a separate retrieval layer and run two systems in parallel.
- At v0.1.0, the nine-phase sleep consolidation pipeline and spreading-activation graph are architecturally novel and community reports on production stability are sparse; teams running high-stakes, continuous-integration workflows who hit an undocumented failure mode in the consolidation cycle have no established workaround and switch to a mature retrieval library such as Mem0 while the issue is resolved.
- The self-hosted sync model depends on an external folder sync tool (Google Drive, Dropbox, git); there is no built-in conflict resolution beyond the memory strength model, so simultaneous writes from two agents on different machines produce a race condition the tool does not explicitly document handling.
About
- Platforms
- npm, CLI, any file-sync service
- API Available
- No
- Self-Hosted
- Yes
- Last Updated
- August 14, 2026
Best For
Who it's for
- Long-term continuity in multi-session coding projects
- Agents that need biologically plausible forgetting and strengthening
- Local-first setups synchronized across devices or cloud folders
- Deterministic identical recall across different models and CLIs
What it does well
- Persistent memory across multiple AI coding agents and sessions
- Retrieval of relevant past decisions under noisy distractors
- Token-efficient loading of only needed skills or facts
- Handling contradictions via decay-weighted recency
- Abstention on missing information to avoid confabulation
Integrations
Frequently Asked Questions
- Is Brain Memory free?
- Brain Memory offers a free option. Hosting, model usage, or optional services may have separate costs; check the pricing details and official terms.
- Is Brain Memory open source?
- Yes. Brain Memory is open source.
- Can I self-host Brain Memory?
- Yes. Brain Memory supports self-hosting on your own infrastructure.
- What platforms does Brain Memory support?
- Brain Memory is available on: npm, CLI, any file-sync service.
Curated lists that include this category
Every time you switch models or start a new session, your agent forgets what it decided last week and you spend the first twenty minutes re-explaining context it already earned.
How Brain Memory works
Brain Memory stores agent decisions as Markdown files with YAML frontmatter, organized in a directory tree you can browse in any file explorer. No opaque vector database or embeddings you cannot audit. Strength decays on an Ebbinghaus exponential curve and rebuilds each time a memory is recalled. The vendor states the tool achieves 100% recall on a 1,000-distractor haystack where BM25 and vector retrievers both score zero. It is at v0.1.0, MIT licensed, and free.
Key capabilities
The docs describe use cases that include persistent memory across multiple AI coding agents and sessions, retrieval of relevant past decisions under noisy distractors, token-efficient loading of only needed skills or facts, handling contradictions via decay-weighted recency, and abstention on missing information to avoid confabulation. Platforms cover npm, CLI, and any file-sync service. Integrations include Claude Code, Codex CLI, OpenCode, Copilot CLI, Kilo, Antigravity, Claude apps, ChatGPT, OpenClaw, Hermes, and MCP connector. A self-hosted option is available.
Strengths and limits
Deterministic recall scoring works across every supported agent and model, so switching from Claude Code to Codex CLI mid-project returns identical memory results. Decay rates differ by cognitive type, letting architecture decisions accumulate strength while one-off notes fade. Plain Markdown files sync via git, Dropbox, or iCloud with no proprietary backend. At v0.1.0 the nine-phase sleep consolidation pipeline and spreading-activation graph remain new, and no public API exists for systems outside the named agent list.
Who it is for / who should skip it
Best for long-term continuity in multi-session coding projects, agents that need biologically plausible forgetting and strengthening, local-first setups synchronized across devices, and deterministic identical recall across different models and CLIs. Teams that need programmatic access for custom agents or high-stakes continuous-integration workflows should skip it until a public API and more stability reports appear.
Contribute
- Work on Brain Memory? Claim this listing.
- Embed a badge for your README or site.
Sign in to rate this tool or suggest an edit.