Pricing
Cloud Free covers solo experimentation with no card. Hosted plans meter on the things that actually cost us money — telemetry events ingested, active memory_blocks in pgvector, projects, seats, retention. Cloud Pro and Team are placeholders: billing isn't wired yet. Join the waitlist and we'll reach out before flipping it on.
Cloud Free
- All 5 hooks captured to telemetry_raw
- Dashboard + pgvector memory search
- JSON export + one-click delete
- Events
- 10,000 / mo
- Memory blocks
- 500 active
- Projects
- 3
- Seats
- 1
- Retention
- 30 days
Cloud Pro
- 250,000 events / mo
- Higher memory + project caps
- 1-year retention
- Events
- 250,000 / mo
- Memory blocks
- 10,000 active
- Projects
- unlimited
- Seats
- 1
- Retention
- 1 year
Cloud Team
- Shared memory_blocks per tenant
- Admin role + audit log
- Pooled event quota across seats
- Events
- 1M pooled / mo
- Memory blocks
- 50,000 active
- Projects
- unlimited
- Seats
- unlimited
- Retention
- 1 year + audit
Need something else?
On-prem deployment or a custom contract? Get in touch — we're in early access and shaping the enterprise tier in conversation with the first teams.
Common questions
What counts as one event?
One row inserted into telemetry_raw. Every UserPromptSubmit, every PostToolUse for a non-filtered tool (Bash | Edit | MultiEdit | Write | WebFetch | Task), every Stop dialog turn, plus the two lifecycle events per session. Read-only tool calls (Read, Grep, Glob, LS) are dropped at the matcher and don't cost you anything.
What's a memory block?
A typed durable fact distilled from your sessions — a single row in memory_blocks with kind ∈ {rule, pattern, decision, …}, a 1536-dim pgvector embedding, and a project_id. 'Active' means visible in your account — deletes count against neither your quota nor your retention window.
Why not meter on API calls instead of events?
One call can carry a 50-token prompt or a 50,000-token transcript. Per-event pricing makes the unit predictable: you know what each hook fire costs and can plan around it.
Why isn't billing wired yet?
We want the metering semantics right before the credit card. Joining the waitlist locks you in at these placeholder rates — or whatever's lower when we flip it on.
What about my Claude API costs?
Contextify doesn't proxy your Claude calls. You bring your own Anthropic key for Claude Code itself; we bring our own key-pool (DeepSeek / Qwen / Gemini) for the extraction worker and the prompt-generation endpoint.