What we shipped.

We pre-announce breaking changes 90 days before the cut. Real release notes start landing here as we ship.

v1.4.2Remote MCP endpoint

Claude Code, Cursor, and Claude Desktop can connect to AgentPrizm over a hosted MCP server with one config block.

v1.4.1AgentSkills marketplace

Private and public skills: install, fork, semantic discovery, maturity flags, and secret/PII scanning before anything ships.

v1.4.0Audit receipts per recall

POST /api/v1/agent/recall now returns each memory's confidence (0–1), filterable with the new minConfidence param, plus an optional receipt that shows why each memory ranked where it did, what it supersedes, and what was filtered (below threshold, deduplicated, low confidence, outside the validity window). Expired memories are now excluded by default — pass includeExpired: true to keep them. All additive; see the API reference.

v1.3.0Validity windows + contradiction handling

Superseded facts decay or are marked stale instead of poisoning new decisions.

v1.2.0Migration from Mem0 / Zep

Move existing memory records into AgentPrizm using REST import patterns.

v1.1.0Claude Code / Cursor quickstart

A five-minute integration path for coding agents over MCP.

v1.0Initial release

The first public version of AgentPrizm. Memory ingest + recall + audit, hybrid semantic + keyword recall, validity windows, and a REST API + MCP server to use them. Quickstart.

Want updates as we ship? Tell usand we'll let you know when material changes land.

Ship agents that remember.

Six lines of code. Confidence scores, validity windows, and audit trails included. Free until your agents ship.

Talk to us