# Give agents, apps, and people the same data and definitions. > A DuckDB lakehouse for people, apps, and agents. Query databases in place, store events, logs, and traces as Parquet, and share metrics across SQL, BI, and MCP. The Context Lakehouse Query data where it lives or bring it into Altertable’s lakehouse. Define what it means once, so every team and tool shares the same context. [Start 14-day free trial](/sign-up) Built for the data you already have Databases & warehouses Query in place Production data via ETL Batch ingestion Events, logs & traces Realtime ingestion ![Altertable Logo](/assets/logos/Logo-full-black.svg) Context Lakehouse - Fast SQL engine - Object storage - Shared knowledge - Access control SQL HTTP API MCP People · SQL & BI Apps · APIs Agents · MCP & tasks Backed by ![20VC](/assets/investors/20vc-white.svg) ![Drysdale](/assets/investors/drysdale-white.svg) ![Kima Ventures](/assets/investors/KimaVentures-white.svg) and angels from ![Algolia](/assets/investors/algolia-white.svg) ![Datadog](/assets/investors/datadog-white.svg) ![Pigment](/assets/investors/pigment-white.svg) ![Sorare](/assets/investors/sorare-white.svg) From stored data to daily decisions ## Your data has more to do than sit in a warehouse. One shared data foundation, useful across the company. Bring your data and business context together so every team can find answers and put them to work. Explore examples by team ### Product ChatGPT @Altertable Where do new players drop off before their first match? 3 lakehouse queries · 2 metric definitions · 4 context entries retrieved **Players drop off after the tutorial.** Only 260 of the 680 players who finish the tutorial start their first match. New-player onboarding · Last 30 days 1. Start game**1,000 100%** 2. Finish tutorial**680 68%** 3. Play first match**260 26%** 4. Return next day**210 21%** 5. 61.8% drop before the first match Check the tutorial-to-match handoff before adding more acquisition spend. Inside the Context Lakehouse ## Keep your stack. Connect the context. Make Altertable your data warehouse, or run it alongside the one you have. Load any table as open Parquet, from production data to events and traces, or query it where it lives. Lakehouse architecture diagram ##### Query in place External catalogs. Data stays in the source. - PostgreSQL - MySQL - MariaDB - MongoDB - Supabase - Snowflake - BigQuery - Amazon Redshift - Apache Iceberg tables - Amazon S3 tables and buckets - AWS Glue Data Catalog - Cloudflare R2 Data Catalog - Hugging Face datasets ##### Ingest Written to Altertable catalogs as Parquet. - Product analytics SDKs - OpenTelemetry logs and traces - Segment events - PostHog: forward captured events with the JavaScript SDK - Fivetran - Airbyte - HTTP API and object storage ##### Context Definitions and docs synced as knowledge. - dbt: models, lineage, and semantic metadata - GitHub repositories - GitLab repositories ##### Query from SQL via Postgres, Arrow Flight, or HTTP. - Hex, through the Postgres adapter - Metabase, through the Postgres adapter - Arrow Flight SQL - DuckDB - Altertable CLI - HTTP query API ##### Agents Any MCP client, with your context built in. - Claude - ChatGPT and Codex - Cursor - Other MCP clients Evaluating Altertable as your data warehouse? See the architecture, deployment options, and performance in detail. [Explore Context Lakehouse](/product/lakehouse.md) Cousteau ## Product investigations on the same foundation. Cousteau watches product events, telemetry, and account data, then investigates what changed, what could explain it, and who is affected. It reads the same catalogs, definitions, and knowledge as your queries and agents. Investigation brief Illustrative - What changed? Product events · semantic model Enterprise export success\_rate fell from 0.987 to 0.912 after 4.12. Other plans moved less than half a point. - What could explain it? Traces · synced runbook Failed export spans cluster near 15 s, the job timeout documented for exports. Enterprise exports are larger on average. - Who is affected? Account records 37 Enterprise accounts had at least one failed export. 12 of them rely on scheduled exports. Each finding cites the evidence behind it. Explanations are leads to verify, not proven causes. Spending hours on recurring investigations across events, logs, and customer data? See how Cousteau runs them. [Explore Cousteau](/product/cousteau.md) Adoption ## What it takes to start. Migration, access, pricing, and how Altertable fits next to the tools you already use. ### Do I need to migrate my data? No. Add existing databases, warehouses, and buckets as external catalogs and query them where they live. To use Altertable as your warehouse, load tables into managed Parquet, from production data to events, logs, and traces. Both can run side by side. Federation avoids copying sources into another raw layer, but Workers still read data over the network and can cache it. ### Why not connect each tool’s MCP server to an agent directly? That works for occasional questions. But each MCP server needs its own credentials and permissions for every person and agent. With Altertable, you connect each system once, then give everyone a single Altertable role that controls which catalogs they can read or write. Agents also get cross-source SQL and shared definitions through one server, instead of piecing them together in each conversation. ### Can I use Altertable with my own agents and tools? Yes. Use SQL, BI tools, the HTTP API, or your own agents through MCP. Cousteau, Altertable’s built-in agent, can work alongside them to monitor your data and investigate changes. ### How does access work? Each person, app, and agent gets one Altertable role that decides which catalogs it can read or write. Agents connected over MCP act for the person who authorized them, and can be kept read-only. To change or revoke access, you do it in one place. ### What do I pay for? Warehouse-grade performance at object-storage economics. Storage and compute are billed separately, with AI tokens billed when you use AI. Start with the free Developer plan, pay for usage with On-Demand, or reserve capacity for volume discounts. ## Latest from Our Blog Stay updated with our thoughts on data, AI, and the tools we deserve. [![Interfaces for the next question](/assets/blog/2026-11-04-interfaces-for-the-next-question.jpg)](/blog/2026-11-04-interfaces-for-the-next-question.md) [NOVEMBER 4TH, 2026](/blog/2026-11-04-interfaces-for-the-next-question.md) [François Chalifour](/blog/2026-11-04-interfaces-for-the-next-question.md) #### [Interfaces for the next question](/blog/2026-11-04-interfaces-for-the-next-question.md) [Generative UI, Architecture, Engineering, Product, Agents](/blog/2026-11-04-interfaces-for-the-next-question.md) [Our UI schema couldn't keep up with new questions, so we built agent-written Data Apps with flexible interfaces and stable data contracts.](/blog/2026-11-04-interfaces-for-the-next-question.md) [READ ARTICLE](/blog/2026-11-04-interfaces-for-the-next-question.md) [![A Semantic Layer in DuckDB](/assets/blog/2026-11-02-semantic-layer-in-duckdb.jpg)](/blog/2026-11-02-semantic-layer-in-duckdb.md) [NOVEMBER 2ND, 2026](/blog/2026-11-02-semantic-layer-in-duckdb.md) [Sylvain Utard](/blog/2026-11-02-semantic-layer-in-duckdb.md) #### [A Semantic Layer in DuckDB](/blog/2026-11-02-semantic-layer-in-duckdb.md) [Engineering, Open Source, Semantic Layer, Analytics](/blog/2026-11-02-semantic-layer-in-duckdb.md) [A semantic layer inside DuckDB: define metrics once in SQL, and every query, from an analyst or an AI agent, gets the joins and aggregations right.](/blog/2026-11-02-semantic-layer-in-duckdb.md) [READ ARTICLE](/blog/2026-11-02-semantic-layer-in-duckdb.md) [![Local Decisions in DuckDB](/assets/blog/2026-10-29-local-decisions-in-duckdb.jpg)](/blog/2026-10-29-local-decisions-in-duckdb.md) [OCTOBER 29TH, 2026](/blog/2026-10-29-local-decisions-in-duckdb.md) [Sylvain Utard](/blog/2026-10-29-local-decisions-in-duckdb.md) #### [Local Decisions in DuckDB](/blog/2026-10-29-local-decisions-in-duckdb.md) [Engineering, Open Source, AI, Analytics](/blog/2026-10-29-local-decisions-in-duckdb.md) [Jev sparked a wave of decision models. We tried a compact open model inside DuckDB, routing support messages with SQL and checking its confident mistakes.](/blog/2026-10-29-local-decisions-in-duckdb.md) [READ ARTICLE](/blog/2026-10-29-local-decisions-in-duckdb.md) ![Altertable Logo](/assets/logos/Logo-full-black.svg) ## Start with one source and one question. Connect a database you already run, or send your first events.\ Query with SQL, build on the API, or connect an agent through MCP. [Start 14-day free trial](/sign-up) Free Developer Plan 10 GB storage · 2 Compute Units · 1M tokens [Explore free plan](/pricing.md#pricing-plans) [Web version](https://altertable.ai/)