Pick Claude Code, Codex, Cursor, OpenClaw, or OpenCode and drop in a scoped MCP key. HTTP and stdio transports both work.
Company context your agents can actually use.
WorkMemory turns the work your team and agents already do into connected, cited, and governed context — then serves it to any agent through MCP. It is a first-class surface of the single Workweaver product: one signup, one governance model, no separate memory plan to buy.
Ask from the agents you already run — Claude Code, Codex, Cursor, OpenClaw, OpenCode, or any MCP-compatible client. Included with Workweaver; pricing is shown after signup or invite.
Your first run
Connect → sync → ask → inspect proofImport AI conversations and agent runs, or stream new work in. Capture and bidirectional proxy expand under their owning workstreams.
Capture roadmap →Your agent queries WorkMemory and gets back cited context — with source conversations and Decision Traces, not an opaque blob.
Open a Decision Trace to see exactly what was recalled and why. Benchmark and latency receipts live on the proof page.
Open proof →Not just a transcript vault
Governance is the differenceContext is isolated across tenant, workspace, team, agent, and user — with a relation graph, not one flat vector store.
WorkMemory promotes raw memories to calibrated facts, precedents, and principles with review and rollback — never blind drift.
Supersession, contradiction handling, and provenance-backed forget are first-class, so the context you trust stays correct.
Every recall can show its sources and reasoning path, so an answer is auditable instead of taken on faith.
Documents, code, images, audio, and video share one ingestion contract where capture exists today — no separate pipeline to wire.
Repeated work surfaces proposed skill improvements you review and approve. Automatic benchmark-driven optimization is in rollout, not claimed yet.
Open WorkMemory inside Workweaver
One product, one signup. Already signed in? You will be handed straight to the app.