Ready with specialized categories
Ships with dbt, Dagster, data analysis, text-to-SQL, RAG, CSV, and more. Ask in plain English. It already knows how those jobs work.
AI & Agents
A general-purpose agent. A custom builder. One Company Brain.
This is where Metroflow shows up in the real world: agents that read the Context Layer, then connect to tools and the ecosystems you already run.
Two ways to run
Start ready-made. Or build your own.
A general-purpose agent for day-one work, plus a custom agent builder when your workflow is specific.
Define the agent you need
Pick tools and capabilities — web search, Slack, Gmail, GitHub, and other integrations — then save it as a reusable agent for your team.
Specialized categories
The general-purpose agent already speaks your stack
Not a blank chatbot. Categories for the work data teams actually do, each grounded in the same Company Brain.
dbt
Dagster
Data analysis
Text to SQL
RAG
CSV
Lineage
Orchestration
Tools & capabilities
Connect agents to the rest of the ecosystem
Give an agent only the tools it needs: search the web, post to Slack, send Gmail, open a GitHub PR. External systems stay outside the data plane. Context still comes from Metroflow.
Web search
Slack
Gmail
GitHub
Calendar
Docs
Grounded in the Context Layer
Company Brain first. Then the tool call.
Every agent — general-purpose or custom — reads the Context Layer before it acts. Certified metrics, lineage, policies, and owners. That is the Company Brain. Slack and GitHub never get a guessed number.
- Same graph Workspace and Context Layer already built
- RBAC: agents see only what your role allows
Automated workflows
Describe the job. Metroflow runs it.
Users define workflows in plain language. The agent uses Company Brain for the analysis, then the tools you attached for delivery.
“Perform data analysis on this dataset every week and send it to my Slack channel.”
Data analysis agent · Slack · weekly
“Compile the certified MRR pack as a PDF and send it to my email.”
Context Layer metrics · Gmail · on close
“If a dbt model fails, open a GitHub issue with blast radius from the graph.”
dbt category · GitHub · on failure
The Metroflow outcome
Workspace and Context Layer, working as agents
Connect the stack. Certify the graph. Then this layer is how that work leaves the product: agents talking to Slack, Gmail, GitHub, and the rest of the ecosystem, still citing Company Brain.
- General-purpose agent for specialized data work
- Custom builder when you need a specific tool mix
- Workflows that run on a schedule or on an event