Sema
semalayer.com
- Free plan
- Web
- API
- Self-hosted
What Sema is
Sema is a semantic layer for enterprise AI that connects databases and answers business questions in plain English with governance. It represents tables, joins, metrics, and business terms in a typed graph; verified inferred joins can be reused in later answers. Sema can ask clarifying questions, show the SQL and graph nodes involved, and choose a chart for an answer. Users can save and schedule email reports and set scheduled alerts for metric thresholds or deviations. Available connectors include PostgreSQL, MySQL, MariaDB, SQLite, and SQLAlchemy URIs; ClickHouse and Amazon Redshift are in beta, while SQL Server, Snowflake, BigQuery, Databricks, and dbt metadata are on the roadmap. It can run in a customer VPC or its own environment, and offers self-hosted Kubernetes deployment using Helm with customer-provided Postgres and object storage. Security features include encryption in transit and at rest, role-based access, restricted-column masking, workspace isolation, and an append-only log. The free Sandbox is limited to one connector, one user, and 100 chat queries per month, with no scheduled delivery or alerts.
Who it suits
Sema is aimed at enterprise and regulated teams, including teams working with warehouses containing thousands of tables. Its deployment and governance options may suit organizations with specific data controls.
Good to know, and what to watch
- Shows SQL and graph nodes used for answers.
- Supports scheduled reports and metric alerts.
- Offers customer-VPC and self-hosted deployment options.
- Includes role-based access and restricted-column masking.
- Free Sandbox permits one connector and one user.
- Sandbox is limited to 100 chat queries per month.
- ClickHouse and Amazon Redshift are in beta.
- Some listed connectors remain on the roadmap.
Plans
Every plan and priceEnterprise
Sandbox
Pro
Business
Free plan · 14-day trial
Getting started
- Open the Sema website
- Select Sandbox, Pro, Business, or Enterprise
- Connect a supported database or SQLAlchemy URI
- Configure users, roles, and identity-provider access
- Use the query interface or connect Sema MCP tools to Claude, Cursor, or an in-house agent
Limits to check
Sandbox includes one connector, one user, 100 chat queries per month, and five saved reports, with no scheduled delivery, alerts, or signals. Pro requires a minimum of three seats, while Business requires at least 10 seats.
The Club's review
Overview
Sema is a semantic layer for enterprise AI that connects databases to business questions in plain English. It is best suited to teams that need governed answers across large or regulated data environments. Its strongest case is the combination of query transparency and deployment controls; its main trade-offs are connector coverage and paid-plan seat minimums.
Key features
Sema organizes tables, joins, metrics, and business terms in a typed semantic graph. It can reuse verified inferred joins in later answers, a useful foundation for teams managing warehouses with thousands of tables. When a question needs clarification, Sema asks follow-up questions; it also shows the SQL it ran and the graph nodes it traversed, then selects a chart for the answer. That makes the output more inspectable than a conversational answer alone, though users still need to assess whether the query answers the business question correctly.
Saved reports can be scheduled for email delivery, and scheduled alerts can flag metric thresholds or deviations. Governed metrics, dimensions and joins, row-level security, and a query API give Sema uses beyond chat. Its capabilities are available as MCP tools for Claude, Cursor, and in-house agents, and it supports OAuth2/OIDC providers including Okta, Entra, Auth0, and Google Workspace.
Available connectors include PostgreSQL, MySQL, MariaDB, SQLite, and any SQLAlchemy URI. ClickHouse and Amazon Redshift are in beta, while SQL Server, Snowflake, BigQuery, Databricks, and dbt metadata are on the roadmap. Teams already standardized on a roadmap connector should wait or assess another option before committing.
Sema says it uses TLS in transit, encryption at rest, role-based access, restricted-column masking, workspace isolation, and an append-only log of queries, plans, and configuration changes. SOC 2 Type II is in progress; GDPR-ready processing terms are available, and HIPAA support with a BAA is available for eligible plans on request. It can run in the customer’s VPC or Sema’s environment; self-hosted Kubernetes deployment uses a Helm chart and customer-provided Postgres and object storage.
Pricing
The Sandbox is free, free forever. It allows one connector, one user, 100 chat queries per month, and five saved reports. Scheduled delivery, alerts, and signals are excluded, and embeds carry “Powered by Sema.” This is enough to explore the product within tight limits, but not to run a shared reporting workflow.
Pro is $49 a month, billed at $49/user/mo, with a minimum of 3 seats and a 14-day trial. It includes unlimited connectors, chat, reports, schedules, alerts, and signals, plus RBAC, an audit log, and email support. The minimum commitment makes Pro a team purchase: the included capacity is broad, but the seat floor makes it a poor fit for an individual or a small two-person group.
Business is $99 a month, billed at $99/user/mo, with a minimum of 10 seats. It includes everything in Pro, SSO and advanced governance (Compliance Agent — V2), AI eval harness and federated joins (V2), priority support, and a 99.9% SLA. The V2 capabilities are marked as such, so teams choosing Business for them should account for that status. The ten-seat minimum and per-user billing point this plan toward larger organizations.
Enterprise has custom pricing, billed annually. It adds SAML/SCIM, a custom DPA, BYOK encryption, an on-prem option using Helm, SOC 2, HIPAA, and GDPR evidence packs, a named customer success manager, and a 99.95% SLA. This is the relevant tier for buyers who need those deployment, identity, or assurance provisions, but its annual custom contract offers no stated entry price.
Platforms
Sema supports API, self-hosted, and web deployment. Its hybrid model accommodates teams that want a hosted service or need to run in their own environment, with self-hosting requiring Kubernetes, Helm, Postgres, and object storage. That flexibility comes with infrastructure responsibilities for self-hosted customers.
Who it's for
Sema fits enterprise and regulated teams that want business users to query data conversationally while keeping semantic definitions, access controls, and query trails in view. Its graph and join reuse are especially relevant to organizations with warehouses spanning thousands of tables. Smaller teams can try the Sandbox, but the one-user cap and paid minimums make it harder to adopt collaboratively; teams dependent on connectors still on the roadmap should look elsewhere for now.
Pros and cons
- Pros: SQL and graph-path visibility make answers easier to inspect; verified joins can be reused across questions.
- Pros: Row-level security, restricted-column masking, audit logging, and customer-VPC or self-hosted deployment address concrete governance needs.
- Pros: Scheduled reports and metric alerts extend the product into recurring reporting workflows.
- Cons: The free Sandbox is limited to one connector, one user, and 100 queries monthly, with no scheduled delivery or alerts.
- Cons: Pro starts at three seats and Business at ten, so small teams pay for a larger team footprint.
- Cons: Several common warehouse connectors, including Snowflake, BigQuery, and Databricks, remain on the roadmap.
Alternatives
Semantic Layer Software is the broader category to explore when comparing semantic-layer products. Choose dbt Semantic Layer if its model-focused approach and plan structure suit your project; its free Developer plan includes one developer seat, 3,000 successful models per month, and one project, with no Semantic Layer listed on that plan. Starburst Enterprise Context Layer is another option to compare; its free plan supports up to three clusters with standard execution for ad hoc queries.
Synmetrix is the alternative to consider if an Apache-2.0-licensed, self-hosted app with no stated paid plan or usage limits is a better fit. Cube offers a free plan with up to five workbooks, 1,000 daily requests, one day of query history, and community support, making it worth comparing when those caps match the intended workload. Kyligence is another paid option with a free trial.
Kyvos is worth comparing for teams considering a cloud marketplace purchase: its listed plan is $0.41 per core hour for the time Kyvos is used. Strata has a free plan with one developer, 25 users, and two data sources per project, a different starting shape for teams testing with multiple users. Definite offers a free tier with two users, two connectors, 1 GB storage, five credits per month, and daily sync; compare it if those limits align better with the initial workload.
Verdict
Sema is a strong fit for enterprise teams that need plain-language data access with inspectable queries, reusable business definitions, and meaningful governance controls. Its deployment choices and reporting features add practical value beyond chat. Choose it when its current connectors cover your stack and the seat minimums suit your organization; look elsewhere if you need a roadmap connector today or a low-cost plan for a small team.
The verdict
Sema connects data structures with natural-language business questions and provides governance controls. Check connector availability and Sandbox limits before choosing a plan.
Questions people ask
Is Sema free?
Yes. The Sandbox plan is free forever and includes one connector, one user, and 100 chat queries per month.
What does Pro cost?
Pro costs $49 a month per user and requires a minimum of three seats; it includes a 14-day trial.
Which databases can Sema connect to?
Available connectors include PostgreSQL, MySQL, MariaDB, SQLite, and any SQLAlchemy URI. ClickHouse and Amazon Redshift are in beta.
Can Sema be self-hosted?
Yes. Sema can run in a customer VPC or its own environment; self-hosted Kubernetes deployment uses Helm with customer-provided Postgres and object storage.
Does Sema support enterprise security and compliance?
It lists RBAC, masking, workspace isolation, and append-only logs. SOC 2 Type II is in progress, GDPR-ready terms are available, and HIPAA support with a BAA is available for eligible plans on request.
What the maker publishes
- Alerts
- Threshold and deviation rules can monitor metrics on a schedule and route breaches to owners with frozen evidence. source
- API
- The roadmap lists a developers API using workspace-scoped keys, and the product states that governed definitions resolve through an API. source
- Audience
- Sema describes itself as built for enterprise and regulated teams, and its About page says it is designed for warehouses with thousands of tables. source
- Audit logging
- Mutations, questions, plans and answers are recorded in an append-only log that can be filtered and exported to CSV. source
- Auto-glossary
- The system drafts business definitions for tables and columns, adds synonyms, sensitivity and lifecycle metadata, and lets people confirm them. source
- Company
- Sema says it was founded in 2025 and is remote-first. source
- Company founding
- Sema's about page states that it was founded in 2025 and is remote-first. source
- Company size and headquarters on LinkedIn
- LinkedIn lists Sema as privately held, headquartered in San Francisco, founded in 2017, with 11–50 employees. source
- Compliance
- Sema’s trust page says SOC 2 Type II is in progress, GDPR-ready data processing terms are available, and HIPAA support with a BAA is available for eligible plans on request. source
- Compliance status
- The company describes DPDP Act alignment, GDPR readiness and a SOC 2 program; its trust page says SOC 2 Type II is still in progress and HIPAA support is available on request. source
- Connectors
- The connectors page lists PostgreSQL, MySQL, MariaDB, SQLite, and any SQLAlchemy URI as available; ClickHouse and Amazon Redshift are in beta, while SQL Server, Snowflake, BigQuery, Databricks, and dbt metadata are on the roadmap. source
- Connectors available
- PostgreSQL, MySQL, MariaDB, SQLite and arbitrary SQLAlchemy URIs are listed as available connectors. source
- Connector status limits
- ClickHouse and Amazon Redshift are beta; SQL Server, Snowflake, BigQuery, Databricks and dbt metadata are listed for the roadmap. source
- Conversational query
- Users ask natural-language questions; Sema can clarify, generate read-only SQL, show the SQL and graph path, and select a chart automatically. source
- Deployment
- Sema says it can run in the customer’s VPC or its own environment; its self-hosted Kubernetes deployment uses a Helm chart and customer-provided Postgres and object storage. source
- Documentation
- The documentation includes a quickstart, sector quickstarts, connector guidance, feature guides, administration, security overview and troubleshooting. source
- Evidence packs
- Each answer can include question, SQL, lineage and access context signed with Ed25519, with expiring read-only audit links. source
- Founded
- 2025 source
- Funding
- Crunchbase shows one private funding round backed by Nimble Ventures, with the amount undisclosed. source
- Identity integrations
- OAuth2/OIDC is designed to work with Okta, Microsoft Entra, Auth0 and Google Workspace, mapping groups to roles. source
- Independent founder profile
- M&A Science describes Matt Van Itallie as Sema's founder and CEO and says the company assesses code quality, technical risk and AI-generated code for software assets. source
- Independent product case study
- Andersen's case study describes an earlier Sema code-review web app with GitHub synchronization, a Chrome extension, reusable snippet search, automatic summaries and tagging. source
- Independent startup profile
- Wellfound describes Sema as an early-stage B2B startup with 11–50 employees and locations in Baltimore and Detroit. source
- Integrations
- Sema describes its capabilities as MCP tools for use with Claude, Cursor, or an in-house agent, and supports OAuth2/OIDC identity providers including Okta, Entra, Auth0, and Google Workspace. source
Checked against the maker's pages on 15 Oct 2026.