Hosted MCP server
Databar MCP: the data layer for your agent
mcp.databar.ai is a hosted MCP server: OAuth, 65+ tools, and 160+ integrations behind them. Claude, ChatGPT, Cursor, or any MCP client can build tables, run enrichments, and push to your CRM. No install, no keys in config files.
14-day full-product trial. No data, no charge.
Launch video
Run your GTM from a chat
Setup
Works with the MCP client you already use
Claude Desktop
Settings → Connectors → Add custom connector → paste https://mcp.databar.ai/mcp
Claude Code
claude mcp add databar --transport http https://mcp.databar.ai/mcp
ChatGPT
Settings → Connectors → Add custom connector → paste https://mcp.databar.ai/mcp
Cursor
Settings → MCP → Add server → https://mcp.databar.ai/mcp
VS Code
Command Palette → MCP: Add Server → HTTP → https://mcp.databar.ai/mcp
Any MCP client
Point it at https://mcp.databar.ai/mcp and approve the OAuth prompt in your browser.
Config-file clients
Windsurf, Cline, Zed and anything else that reads mcpServers: { "mcpServers": { "databar": { "url": "https://mcp.databar.ai/mcp" } } }
Create, iterate, and deploy - 65+ tools, 450+ enrichments included
Tables
Create tables and columns, insert rows or push a CSV straight in, and read results back with filters.
Search
Find companies and people by filter before you have a single row, then search the catalog by name.
Enrichments
Run 450+ enrichments across 100+ data providers, on one row or the whole table.
Waterfalls
Sequence providers per field and pay only for verified answers.
Flows
Build, version and validate multi-step flows — and estimate what a run costs before it starts.
Sources and exports
Schedule syncs, wire up webhooks, and push finished rows to your CRM or CSV.
Beyond one-shot lookups
Waterfalls and flows, driven from the chat
Most MCP servers hand an agent one lookup at a time. This one gives it the same machinery the app runs on.
→ Nine email providers in sequence, 3 to 22 credits, first hit wins
→ Emailable, Bouncer or ZeroBounce verification for one more credit
→ Fifteen flow tools: build, validate, version and price a run
Live sources
Attach a webhook or a scheduled source to a table, map its fields, then pause, resume or re-sync it on demand.
Bulk and CSV
Push a CSV straight into a table, or run enrichments, waterfalls and flows in bulk across every row at once.
Run control
Check task status mid-run, and cancel a table enrichment or export that is heading the wrong way.
Security and control
Secure and sandboxed by design
OAuth, not a key in a text file
Connecting runs an OAuth handshake. Nothing secret goes into a config file you might commit or paste into a chat. Credentials you add inside Databar are never exposed to third party agents.
Scoped to your workspace
The agent sees what your own account sees - the same tables, credits and connected integrations. Nothing wider.
Every run is inspectable
Tool calls land as real tables and runs in the app, so you can open anything an agent did and check it by hand.
Build a complete system from a single prompt
Qualify inbound while you read it
Paste the domains from overnight signups, get size, funding and stack back in a table, then push the keepers to your CRM.
Build a list without leaving the chat
Search companies and people on the same filters as the app, enrich in a waterfall, export once it looks right.
Keep the CRM from going stale
Have the agent re-run enrichments on accounts nobody has touched in 90 days and write the fresh fields back.
Same credits, same refunds
Agents get no special meter. Failed lookups cost nothing, whether a human clicked or a model called a tool.
No agent surcharge
Tool calls are metered on the same price list as clicks in the app.
Empty results refund themselves
When a provider comes back with nothing, the credits go back — no ticket, no review.
The agent can check first
Balance and per-enrichment cost are tools of their own, so a model can price a run before starting it.
Questions
The things people ask before connecting
Do I need an API key?
No. Connecting runs OAuth in your browser and the client keeps the token. Nothing secret goes into a config file.
Which clients work?
Anything that speaks MCP over HTTP — Claude Desktop, Claude Code, ChatGPT, Cursor, VS Code, Windsurf, Cline, Zed, or an agent you wrote yourself.
Does it cost more than using the app?
No. Tool calls draw on the same credits at the same rates, and failed lookups are refunded either way.
Can an agent spend my whole balance?
It can only spend what your workspace holds. Balance and per-enrichment cost are tools of their own, so a model can price a run before it starts.
Where does the data end up?
In real tables in your workspace. Open, edit and export them by hand, or push them to your CRM — the agent is not a separate silo.
Is there an API too?
Yes. A REST API, a Python SDK and a CLI cover the same ground if you would rather write code than chat.
