hivemind

Activeloop's shared brain for agent TEAMS: traces from Claude Code, Codex, Cursor & co become reusable skills every teammate's agent can execute — cloud-backed, 25% cheaper on LoCoMo.

1,505 94 TypeScript Apache-2.0updated 3 days ago

memory coding

View on GitHubDispute this mapping →

Curator's take

The pitch is the org-level version of agent memory: one engineer's agent figures out the tricky migration on Monday, every agent on the team executes the pattern Tuesday. Auto-learning from traces across seven agent hosts, with real LoCoMo receipts (25% cheaper, 1.7x fewer tokens vs no shared memory). Reach for it when the unit of learning is the team, not the seat. NOT local-first: cloud-backed on Deeplake is the architecture AND the business model (Activeloop, YC) — traces of your engineers' sessions leave the machine, so clear it with whoever owns your IP policy before the whole team wires in.

Mapped by ShipWithAI editors · links verified

README.md

Hivemind

One brain for all your agents

Auto-learning, cloud-backed shared brain for Claude Code • OpenClaw • Codex • Cursor • Hermes • pi • Claude Cowork (Alpha) agents.

One engineer's agent figures out a tricky migration on Monday.

Tuesday, every agent on the team can execute the pattern.

On LoCoMo, the public long-context memory benchmark, Hivemind is 25% cheaper, 1.7× fewer tokens, and 31% fewer turns than running without shared memory. (See the numbers below.)

Cut Cursor spend by 34% with Hivemind

Beyond memory. Hivemind doesn't just remember. It mines your team's traces for repeated patterns and codifies them into reusable skills that propagate back into every agent on the team. The agent your junior engineer used this morning is sharper because of what your senior engineer's agent figured out last week.

📥 Captures every session's prompts, tool calls, and responses as structured traces in Deeplake
🧠 Codifies patterns into reusable SKILL.md files, available to every agent on your team
🔍 Searches traces and skills with hybrid lexical + semantic retrieval (BM25 fallback when embeddings off)
🔗 Propagates capability across sessions, agents, teammates, and machines in real time
📁 Intercepts file operations on ~/.deeplake/memory/ through a virtual filesystem backed by SQL
📝 Summarizes sessions into AI-generated wiki pages via a background worker at session end
☁️ BYOC: keep data in your own GCS, Azure, S3, or on-prem bucket. See Security & storage

Benchmarks

On the LoCoMo long-context memory benchmark (100 QA pairs, Claude Haiku via claude -p, hybrid lexical + semantic retrieval), Hivemind cuts cost, tokens, and turns versus a no-memory baseline:

Metric	Baseline	Hivemind	Improvement
Cost / 100 QA	$8.94	$6.65	25% cheaper

Continue your stack

What teams reach for next — and why each earns a place beside hivemind. Ranked by curator confidence.

Alternative to See all + compare →

memsearch

Zilliz's unified memory for coding agents: one Markdown + Milvus store shared across Claude Code, Codex, OpenCode and OpenClaw — hybrid search, plus repeated workflows distilled into skills.

Why it fitsBoth distill agent sessions into shared, reusable skills across multiple agent hosts. memsearch is local-first, one machine, Markdown+Milvus; Hivemind is cloud-backed and team-scoped — your colleague's agent learns from yours.

agentic-context-engine

Learning loop for any agent: reflect on failures, distill strategies into a Skillbook, inject them next run — 2x consistency on Tau2, 49% token cuts. LiteLLM-based, 100+ providers.

Why it fitsSame learn-from-experience loop, different scope: ACE is an embedded Python engine distilling strategies for the agent you're building; Hivemind is a service layer distilling skills across every off-the-shelf agent your team runs.