Every paid plan includes the complete Sift product. Choose only how much customer-visible source storage and pooled hosted compute your workspace needs.
Start with the full collaborative brain.
per seat, billed annually
More retained source context and compute.
per seat, billed annually
The largest self-serve memory and compute pool.
per seat, billed annually
A considered rollout for organizations with more moving parts.
One product on every paid plan. Source storage is per workspace; hosted-compute credits are granted per seat and pooled for the team.
Connect
Roadmap notes, customer calls, GitHub work, and launch decisions enter one permissioned brain.
Illustrative rates. Actual usage settles from measured work.
30,800 credits remain
58 GB remain
This illustrative month fits Scale with room to keep working.
Yes. Every paid plan includes integrations, collaboration, permissions, Skills, hosted agents, MCP, CLI, and API access. Plans change capacity, not capability.
Start with the shape of a normal month. Use the simulator above, then choose enough storage and pooled credits for the team. You can revisit the plan as the workload becomes measurable.
When the rollout needs a conversation before a checkout. Enterprise is for larger workspace planning, procurement or security review, and deployment coordination. Terms are defined with you, not implied on this page.
Yes, when real work begins. The cardless 14-day Launch trial starts when Sift first stores a source or accepts hosted compute, with 25 GB, 5,000 pooled credits, and up to three active people.
It depends on the work mix. The simulator gives a transparent planning estimate across agent runs, indexing, learning, images, documents, and media. Your ledger records actual measured usage.
The underlying provider work differs. A multi-call agent run, source indexing, a learning exercise, and image understanding do not cost the same to operate, so Sift records them by category.
Your existing brain stays available. Hosted compute or new source writes pause at the relevant limit; read, search, citations, deletion, and export continue. Sift does not silently delete stored work.
Included credits last one billing cycle. Purchased credit packs remain available for 12 months, giving teams a longer buffer for exceptional workloads.
Access boundaries apply before retrieval. Workspace, source, and connector permissions determine what people and agents can see; personal connectors remain private unless their visibility is changed.
No. Sift does not use your workspace content to train generalized AI models.
Paid access continues to the billing boundary. The workspace then becomes read-only so owners can export or delete data and recover billing without losing access to existing work.
Yes. Workspace owners can export or delete workspace data. Limited operational, audit, and backup records may remain for the periods described in Sift's privacy policy.
Sift is the shared context layer for your company. People and agents work from the same permissioned, cited knowledge, and every outcome improves the next run.