gbrain vs hindsight
Two of the top projects in Memory, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
gbrain
Long-term memory store for coding agents, with sources and corrections
hindsight
Agent memory server with retain, recall and reflect operations
Verdict
gbrain ranks higher, 81 to 78. Of the 5 score parts we measure for both, gbrain leads on 2, hindsight on 1 and 2 are even.
- Adoptionhindsight leads by 36 points, 77 to 41: a wide gap.
- FreshnessEven at 100.
- Maintenancegbrain leads by 4 points, 98 to 94: close to even.
- Easy to rungbrain leads by 33 points, 83 to 50: a wide gap.
- Agent-readyEven at 85.
Pick gbrain if you are
Developers giving Claude Code, Codex or other agents persistent shared memory
- Keyless start: PGLite brain with keyword search, no server or Docker needed
- One brain shared across agents via stdio MCP or HTTP MCP with scoped OAuth tokens
Pick hindsight if you are
Developers adding long-term memory to LLM agents and coding assistants
- Four parallel retrieval strategies merged with reciprocal rank fusion and cross-encoder reranking
- Works with 25+ LLM providers, including local ollama, lmstudio and llamacpp
| What we compare | gbrain | hindsight |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 41, known | 77, popular |
| Freshness | 100, active | 100, active |
| Maintenance | 98, healthy | 94, healthy |
| Easy to run | 83, very easy | 50, easy |
| Agent-ready | 85, ready | 85, ready |
| Facts from GitHub and the README | ||
| Stars | 30.9k | 49.7k |
| License | MIT (permissive) | MIT (permissive) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Oct 2026 | Oct 2026 |
| Language | TypeScript | Python |
| Docker | Yes | Yes |
| GPU | Not needed | Not stated |
| arm64 or Apple Silicon | Mentioned | Mentioned |
gbrain
GBrain keeps explicit facts and their sources in a local PGLite or Postgres brain and exposes them to agents such as Claude Code, Codex, Cursor and Hermes over stdio or HTTP MCP. It starts keyless with keyword search; semantic search, cited synthesis, a typed knowledge graph and an overnight enrichment cycle are optional. It is installed with Bun from GitHub, not npm.
Who it is for: Developers giving Claude Code, Codex or other agents persistent shared memory
Strengths
- Keyless start: PGLite brain with keyword search, no server or Docker needed
- One brain shared across agents via stdio MCP or HTTP MCP with scoped OAuth tokens
- One-command setup for Claude Code and Codex, with dry-run and remove options
- Optional cited synthesis that notes gaps, plus a typed graph for relationship queries
Weaknesses
- Not on npm; the unrelated npm package named gbrain can shadow the real binary
- Always-on OpenClaw/Hermes setup needs a server with 8GB+ RAM plus paid API usage
- Local files and shared database credentials are not isolated by sources
- Remote put_page does not extract graph edges inline; HTTP needs explicit maintenance
- RAM ≥ 8 GB
- no GPU
- Compose
- Needs Bun 1.4.0+, PGLite or Postgres, Tailscale (optional, for remote MCP)
- Models: Optional cloud embedding, reranking, extraction and synthesis providers (unspecified)
hindsight
Hindsight stores agent memories in banks and extracts facts, entities and timestamps from retained text using an LLM. Recall runs semantic, BM25, graph and temporal retrieval in parallel, then reranks the merged results. Background jobs consolidate facts into observations and mental models. It exposes a REST API, Python, Node.js and Go clients, a CLI, and a per-bank MCP endpoint.
Who it is for: Developers adding long-term memory to LLM agents and coding assistants
Strengths
- Four parallel retrieval strategies merged with reciprocal rank fusion and cross-encoder reranking
- Works with 25+ LLM providers, including local ollama, lmstudio and llamacpp
- Built-in MCP endpoint per bank, plus 60+ listed integrations
- Embedded mode runs in-process via pip with a bundled pg0 database
Weaknesses
- Every retain call requires an LLM, adding cost and latency
- Accuracy claims are the vendor's own benchmarks; independent reproduction is partial
- Managed Cloud and Enterprise tiers exist; feature differences from self-hosted are unclear
- Minimum RAM and GPU needs are not stated in the README
- Docker
- Needs LLM provider (hosted or local), PostgreSQL (embedded pg0 by default), Oracle AI Database (optional)
- Models: openai, anthropic, gemini, groq, bedrock
- port 9999
More in Memory
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.