Authenticated Access

Persistent memory for serious operators.

WorkMemory pricing is shown only after sign-in. During guided rollout, pricing is revealed after account creation so we can qualify fit, apply invite rules correctly, and keep the product surface aligned with the actual offer.

`@bitfoundry.ai` accounts receive complimentary access during rollout. Invited users see their catalog after sign-in. Everyone else joins the waitlist first.

What unlocks after sign-in

No public price exposure during rollout
API + MCP memory Search, recall, import, pin, forget, and session memory from one product surface.
Multimodal ingest Documents, code, images, audio, video, and connector-fed attachments share one ingestion contract.
Tenant + workspace isolation Context is scoped across tenant, workspace, team, agent, and user boundaries.
Evidence-backed evolution Patterns, precedents, and decision trees compound with review and rollback, not blind drift.

`@bitfoundry.ai` access

Complimentary during guided rollout. No card. No checkout. Create the account and the internal bypass applies automatically.

Complimentary rollout access

Invited users

Create the account first. Authenticated pricing, bundled tiers, and invite benefits are revealed inside the product after authentication.

Invite rules apply automatically

Everyone else

Join the waitlist. We are optimizing for fit, feedback quality, and runtime reliability before broader rollout access.

Waitlist before broader rollout

Why we gate the offer

Guided rollout, not price-led signup
1
Verify fit We want the first rollout users to match the current product shape: agent builders, operators, and teams who need persistent context more than a thin chatbot wrapper.
2
Reveal the right catalog Workweaver bundles WorkMemory differently by plan, invite discounts vary, and internal accounts run complimentary. The post-signup surface can tell the truth cleanly.
3
Keep the public story simple Public landing should sell the product thesis and access model. Authenticated product should carry pricing, billing, and plan detail.

Current proof points

Public capability signal, not public price list
8.9ms p50 recall latency

Measured on the current committed regression artifact, not a synthetic landing-page estimate.

94.63% Internal MemoryBench regression

145-query regression artifact with tenant isolation and evidence-aware memory contracts.

4-layer Scoped memory architecture

Hot, warm, cold, and pattern layers working together instead of one flat vector store.

See /memory/proof.html for the current artifact path, benchmark breakdown, comparison matrix, and explicit claim boundaries.