tools to answer one question
Slack → BI → spreadsheet → SQL editor → docs → back to Slack. The round trip kills momentum and erodes trust.
Built with our design partners across —
Every dashboard hides a maze of joins, business rules, and tribal knowledge. AI on top of that maze guesses — confidently and incorrectly.
Slack → BI → spreadsheet → SQL editor → docs → back to Slack. The round trip kills momentum and erodes trust.
Revenue, churn, MAU — defined three different ways in three places. Nobody knows which is right anymore.
LLMs hallucinate against undocumented schemas. Without a semantic layer, retrieval grabs the closest-named field, not the right one.
Tables, columns, metrics, and business terms live as a typed graph — not a wiki, not a YAML file. Every query, dashboard, and agent reads from the same source of truth, and changes propagate.
Conversation is the input; a versioned canvas of charts, queries, and explanations is the output. No more screenshotting numbers from one tool into another.
Row-level access, column masking, query logging, and signed answers are on at install. Turning governance off is a deliberate choice — not the consequence of forgetting it.
From source connection to audit-grade answer — every layer of the semantic stack ships today.
40+ sources, sandboxed sync.
Live introspection on connect.
Freshness, volume, nulls at a glance.
Plain-English definitions, generated.
Plain English in. Auditable chart out.
Save, schedule, and deliver.
Know the moment a metric moves.
Drift becomes a governed record.
Verified joins, first-class metrics.
Role-appropriate access, enforced.
Append-only, compliance-grade.
Governance isn't a feature bolted on the side — it's woven through every answer Sema returns. Here's the machinery that makes an answer defensible.
Ed25519 signatures on every answer.
Each answer is bundled with its question, SQL, lineage, and access context, then cryptographically signed. Re-verify the signature any time — tampering is mathematically detectable.

Read-only, expiring, no login.
Send an auditor a link that renders the signed pack — SQL, lineage, signature badge, PDF export — with zero access to the rest of your workspace. Every open is logged.
DPDP, BFSI & SaaS check sets.
Curated, industry-specific compliance checks run on a schedule and freeze evidence the moment they fire — audit-readiness on autopilot.
Verify a join once. Answer forever.
Tables, columns, metrics, and business terms live as a typed graph. Confirm an inferred join once and every future answer inherits the correct path.
Drift becomes an immutable record.
Threshold and deviation alerts on any metric. A reviewed breach becomes a signal with provenance in an append-only registry — prove what you knew, and when.
Plain English in, auditable chart out.
Ask in your own words. Sema asks clarifying questions before guessing, shows the SQL it ran and the nodes it touched, and picks the right chart automatically.

No pipelines to rebuild, no metrics to re-model by hand. Sema meets your stack where it is.
Point Sema at your warehouse or upload a file. It introspects the schema, profiles every table, and maps the terrain in seconds — sandboxed and read-only.
Sema drafts plain-English definitions, infers joins, and proposes metrics. You confirm what's right; the semantic graph becomes your single source of truth.
Anyone on the team asks in plain English and gets a governed, explainable answer — with the SQL, lineage, and a signed evidence pack attached.
Anyone can wire an LLM to a database in an afternoon. What's hard — and what Sema does — is everything between the question and an answer you can defend: understanding meaning, planning safe SQL, visualizing it, and signing the result. Six layers, working as one.
beyond a chatbot
Sema picks the right chart for the shape of the answer — bar, line, choropleth, stat — and renders it automatically. No dragging fields onto a canvas.
10+ chart types, auto-selected
Ask a broad question and Sema plans the sub-queries, assembles the findings, writes the narrative from real numbers, and builds a shareable deck — PPTX, PDF, or web.
Question → full analysis & deck
Row-level access, column masking, refusals, and Ed25519-signed evidence are on by default — not a plugin you remember to add later.
Signed & audited, every answer
Collapse the Slack → BI → spreadsheet → SQL round-trip into one prompt. Audits that took weeks take an afternoon; analyst hours go back to analysis.
Weeks of audit prep → hours
Each layer is months of specialised engineering. Together they're the reason an answer is trustworthy — and the reason this is hard to replicate.
by the numbers
built for your stack
The same governed semantic layer, tuned to the questions your sector actually asks. Pick yours.
Readmissions, occupancy & CMS reporting — answered, with an audit trail.
ExploreChurn, NRR, MAU & pipeline — one definition the whole company trusts.
ExploreRisk, exposure & regulatory reporting — explainable and fully audited.
ExploreSales, inventory, margin & basket — answered across every channel.
ExploreSema runs in your VPC or ours. The semantic graph and the execution sandbox are separable — keep your data plane in your network.
Every Sema capability is an MCP tool. Drop the toolbox into Claude, Cursor, or your in-house agent and inherit Sema's governance.
Bring your IdP — Okta, Entra, Auth0, Google Workspace. Group-mapped roles flow into row-level policies automatically.
Query plans run in ephemeral containers with no outbound network. Generated SQL is statically analyzed before it ever sees a database.
Helm chart for K8s. Bring your own Postgres + object store. Sema's control plane never holds your data — only the graph.
Sema is built for regulated teams — aligned to India's DPDP Act 2023, GDPR-ready, and running a SOC 2 program. Your data plane can stay entirely inside your network.
Purpose limitation, consent records, data-principal rights workflows, and breach-notification readiness built into the platform.
TLS 1.2+ in transit, AES-256 at rest. Workspace LLM keys are Fernet-encrypted; secrets never touch the client.
Viewer, analyst, admin, owner — enforced in code. Restricted columns are refused or masked at plan time, before any SQL runs.
Every mutation and every answer recorded immutably, filterable by actor, kind, and date, exportable for your auditors.
Deploy Sema's control plane in your own Kubernetes. Bring your own Postgres and object store — Sema never warehouses your rows.
SOC 2, HIPAA, and GDPR evidence packs generated from the live audit trail — not assembled by hand the week before a review.
Start free in a sandbox. Upgrade to a 14-day Pro trial — no card to start — and convert when your team is hooked.
Evaluate Sema against your own data, free.
min 3 seats · 14-day trial
For teams running on governed answers.
min 10 seats
For orgs that need SSO and advanced governance.
For regulated, multi-region, and self-hosted deployments.
We build in the open with a small cohort of design partners across regulated industries.
"Sema turned our 'which number is right?' Slack threads into a single governed answer with the SQL attached. Our audit prep went from weeks to an afternoon."
"The signed evidence packs are the unlock. When compliance asks how we got a figure, we send a link — signature, lineage, and all."
"We connected our warehouse on a Friday and the whole ops team was self-serving governed answers by Monday. No dashboards to rebuild."
Tell us what you're trying to answer. We'll show you Sema running against a dataset like yours — and reply within one business day.
Everything you need to write about Sema — logos, brand colors, a company one-pager, and product screenshots. Grab an individual asset or the full kit.
Sema is the semantic layer for enterprise AI — governed, explainable text-to-SQL that lets teams ask business questions in plain English and get audit-ready answers with cryptographically signed evidence.
Press enquiries: press@datafrontier.co