Hunta Gather
Attestable, curated memory for AI agents and teams. Verify, don't trust.
Hunta Gather is a memory service for AI agents with two properties raw memory stores do not have:
- A curated write path. Agents propose; nothing enters canon by fiat. Every write passes a screening and judging gate before it becomes a sealed, attributable fact. The writer never decides.
- Provable isolation. Every tenant can request a signed attestation of its isolation boundary and verify it offline against Hunta's published key. The verifier is open source, so there is nothing to take on faith.
Recall returns only sealed facts, with provenance. Time is bi-temporal: what was true, and when you learned it.
Start here
Quickstart
Get a key and make your first write and recall in a few minutes.
Verify the attestation
Prove your store's isolation boundary in about 60 seconds.
Connect over MCP
Point any MCP client at your memory.
Claude Code plugin
Auto-recall and loss-proof capture on every session.
The shape of the system
- Gather is the memory layer: capture, curation, recall.
- Instinct is the governed reflex set: standing constraints your agents consult every turn.
- Tribe is shared team memory: one attested store, many seats, per-seat access control.
The engine is a hosted service at mcp.hunta.ai (an MCP server plus a REST API). The client
integrations and the attestation verifier are open source at
github.com/hunta-ai/gather.