Free for solo use.
You pay when you share.
Use FeatureQL alone with pip, or with the managed registry up to 2,000 queries a month. Teams pay only for the people who use it.
PIP
The Python library with a local registry, for one person. No account, nothing sent to FeatureMesh.
- The full FeatureQL language
- All supported databases
- MCP server for AI assistants
- Embedded real-time serving
Managed
The managed registry in FeatureMesh Cloud. Occasional team members are free.
Everything in PIP, plus
- Shared catalog with revisions and branches
- Datakey caches
- Permissions and feature lifecycle
- Rust serving engine: 1,000 inferences count as 1 query
- Analytics proxy for your SQL clients
Custom
For deployments across the company, close to your data and applications.
Everything in Managed, plus
- Self-hosting on your infrastructure
- Contract and invoicing
- Consulting to integrate with your stack
- Dedicated support
hello@featuremesh.com
What the managed registry adds
PIP is for one person. The managed registry is where definitions become shared, versioned and governed.
| Capability | PIP | Managed | Custom |
|---|---|---|---|
| Language | |||
| FeatureQL, alone or inside SQL Typed, composable definitions that compile to standard SQL. | |||
| Several databases The same definitions on DuckDB, Trino, BigQuery, Snowflake and more. | |||
| AI assistants A local MCP server for validation, diagnosis, execution and help. | |||
| Catalog | |||
| Shared catalog Teams discover, reuse and compose each other’s features. | Single user | ||
| Catalog revisions Every change is logged. Read the catalog as it was with AS OF. | — | ||
| Branches Try a change without touching live definitions, then apply it. | — | ||
| Datakey caches Queries read precomputed tables when it saves work, never stale ones. | — | ||
| Governance | |||
| Access management Orgs, projects, roles, invitations and service accounts. | — | ||
| Namespace permissions Decide who may change which part of the catalog. | — | ||
| Feature lifecycle DEV to DEPRECATED, enforced across the dependency graph. | — | ||
| Deployment | |||
| Real-time serving Features as APIs, from Redis, JDBC and HTTP sources. | Embedded | Rust engine | Rust engine |
| Analytics proxy Existing SQL clients send FeatureQL through an on-premises proxy. | — | ||
| Registry hosting Where the managed registry runs. | Your machine | FeatureMesh Cloud | Managed by us or self-hosted |
Managed covers the Personal and Team plans. Click a capability to read its documentation.
Estimate a team's month
Count only the people who query. Occasional users do not add to the bill.
Questions
Does FeatureMesh see my data? +
No. FeatureMesh holds definitions, not rows. It translates FeatureQL to SQL, and that SQL runs on your databases.
What counts as a query? +
A FeatureQL query, or SQL with embedded FeatureQL, that the managed registry translates. On the serving engine, 1,000 inferences count as one query. Failed queries, plain SQL and everything you run with pip are free.
Who is an active member? +
Anyone, person or service account, who ran 10 queries or more in the month. Occasional readers cost nothing.
What if we need more than 2,000 queries per member? +
Extra queries are pooled across the team and cost $20 per 2,000.
How do we pay? +
Buy credits by card in the console, once or as a monthly reload. Custom contracts are invoiced.
Can we move from pip to the managed registry later? +
Yes. The same definitions move to the managed registry without a rewrite, and every definition compiles to standard SQL you can keep if you leave.
The full billing rules are in the documentation.
Deploying across the company?
Talk to us about self-hosting, contracts and help with the integration.
Contact us at hello@featuremesh.com