Agents Schema plugin
A plugin that connects an agent to a data warehouse and finds table, dbt model and semantic object metadata in the AGENTS schema.
- 72 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The connection skill configures agent access to the warehouse and runs SQL.
- Workflows write the schema to the warehouse with an account from a GitHub secret.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/agents-schemaDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The connection skill configures agent access to the warehouse and runs SQL.
- Workflows write the schema to the warehouse with an account from a GitHub secret.
This is third-party code. Review the repository files before installing.
What it does
Agents Schema puts context about data inside the warehouse itself: a standard `AGENTS` schema holds table descriptions, owners, dbt model sources, LookML and Omni views and OSI semantic models. GitHub Actions workflows from the repository load this metadata into Snowflake, Databricks, BigQuery or ClickHouse. The agents-schema plugin for Claude Code and Codex installs two independent skills. The `connect-warehouse` skill configures and verifies warehouse access with a test query, without asking for passwords or tokens in chat. The `agents-schema-search` skill, once connected, searches metadata through `AGENTS.ROOT`: which tables exist, how they are described, whether they are fresh and who owns them.
Who it is for. Analysts and data engineers whose agent writes SQL against a warehouse and needs context about tables.
Good fit when
- An agent answers a data question and must find which tables fit.
- Claude Code or Codex must be connected to Snowflake, BigQuery, Databricks or ClickHouse.
- dbt project metadata is published to the warehouse automatically after each build.
- You need to know who owns a dataset and whether its source is fresh.
Not a fit when
- An agent edits dbt sources and works with project files and artifacts.
- The warehouse is not among Snowflake, BigQuery, Databricks and ClickHouse.
- An application needs a rigid deterministic contract, while the schema is designed for an agent reader.
Example request
Connect to our Snowflake and find revenue tables in AGENTS.ROOT: who owns them and when the data was last refreshed.Limitations
Workflows write to the warehouse with credentials from a single GitHub Actions secret, `WAREHOUSE_CREDENTIALS`, so that account should have minimal rights to write the schema. ClickHouse has separate setup notes. The plugin skills run locally and publish nothing. The `agents-schema-analyst` row in `AGENTS.ROOT` remains warehouse-side behaviour. Connecting to a warehouse requires your own access to it, and queries consume its compute resources.
How to disable. Run `claude plugin uninstall agents-schema@agents-schema` and optionally remove the marketplace. Drop the `AGENTS` schema from the warehouse and the GitHub secret separately.
FAQ
How do I install the plugin in Codex?
With `codex plugin marketplace add dbt-labs/agents_schema` and `codex plugin add agents-schema@agents-schema`, then start a new task.
Which skills does the agent get in Claude Code?
`/agents-schema:connect-warehouse` configures access and `/agents-schema:agents-schema-search` finds metadata. Claude may also invoke them itself.
Where does metadata come from?
From dbt, Looker, Omni, OSI, Sigma and Markdown skills. dbt is fastest: the project path and `target/manifest.json` are enough.
Which warehouses are supported?
Snowflake, Databricks, BigQuery and ClickHouse. Each setup guide has a sample credential shape.
Does the schema replace a dbt MCP server?
No. Editing a dbt project still needs its files and artifacts, while the schema is a shared queryable metadata surface.
Discussion
No questions or reviews yet.
Related
PluginsThe Dex plugin: an agent explores a database, builds dbt models and a semantic layer and tracks schema drift under cost control
AWS hcls-agents plugin gives Claude Code and Codex skills for genomics, clinical trials and biomarkers plus MCP servers.
A database CLI for dozens of engines with an agent mode: JSON output, a read-only guard, limits and a broker that hides passwords
Neo4j skills for agents: Cypher queries, graph modeling, drivers, import, GraphRAG, GDS and Aura provisioning
A Google Cloud plugin catalog for databases and analytics: AlloyDB, Spanner, Bigtable, Firestore, Cloud SQL and Looker.
A template for data scientists: AGENTS.md, CLAUDE.md and skills for Python, SQL, Polars and model diagnostics.
README in short
The repository holds workflows, a plugin marketplace for Codex and Claude Code, the `SPEC.md` schema specification and examples. The idea is that an agent is the metadata reader, so providers can store data semi-structured and change shape without agreeing a rigid contract.