Pricing
We do not meter the behaviour we want to encourage. Pick the plan that fits how your team runs agents today — memory operations stay unmetered for normal use.
Builder
Free
Free for one person. A private workspace of your own.
For one person proving the loop on their own work — memory that is private to you, in whatever tool you open next.
includes
Your own memory, private by default
Both domains — coding and knowledge work
Works in Claude Code, Codex, Cursor, Copilot
Memory operations unmetered for normal use
Community support · cloud only
Teams
Free
Up to 4 people · then $599 per user / year
For engineering teams that want AI work to compound across people, repos and tools.
everything in builder, plus
Private team memory across repos, tools and agents
Decisions, fixes and patterns captured once, reused
Same memory in Claude Code, Cursor, Codex, Copilot and custom agents
Review, edit, deduplicate and retire shared memories
Memory reuse, agent activity and token-savings insights
free for up to 4 people · paid seats billed annually, saving 27% vs $828 monthly
Enterprise
Custom
Platform and programme licensing.
For organisations running agents across teams, tools, models and governed workflows.
everything in teams, plus
SSO / SCIM, RBAC and audit logs
Policy-seeded memory and provenance
Client / programme memory scoping
VPC, on-prem and UK/EU residency
Dedicated support and paid pathfinders
No metering on memory
No charge per memory call, retrieval or agent run. If memory is working, your agents should use it constantly.
Priced per contributor
You budget around people, so do we. Production-agent scale is Enterprise licensing, not surprise usage fees.
Your learning, owned
What your agents discover becomes permissioned, auditable organisational memory that survives model and tool churn.
Questions, answered
Tell us how your team runs agents and we’ll point you to the right plan — or scope a pathfinder if you’re bringing Spark into an enterprise programme.
the loop
benchmarks · product · research
A short dispatch on shared memory for AI agents — the numbers behind the product, what we're shipping, and the research we're reading. No filler.