Ask your AI anything about the company & job market.
Connect Jobo to Claude, ChatGPT, Cursor, or any MCP client and let your agent run live analytics over the entire company + job dataset—funding, hiring, tech stacks, comp, layoffs, talent—in plain English. It writes the SQL; you get the answer. No data to pipe, no pipeline to build.
https://mcp.jobo.world/mcpWhat you can ask.
Your agent turns plain-English questions into SQL over the full company + job graph—no schema to learn, no exports to wrangle. A few of the questions teams ask every day:
Sales & GTM
Technographics + buying intentInvestors · VC & PE
Deal sourcing + diligenceRecruiters & talent intel
Comp, skills & supplyAnalysts & researchers run the same two tools for labor-market and sector trend reports—every answer is live, sourced, and reproducible.
Two tools, the whole dataset.
The MCP server exposes just two tools—your agent learns the schema once, then runs read-only DuckDB SQL against a daily snapshot of every company and job. Built for aggregation: GROUP BY, percentiles, and joins across funding, hiring, skills, and comp.
Describe
Your agent calls describe_dataset to learn every table, column, join key, and the real distinct values—so it writes correct SQL on the first try.
Ask
You type a plain-English question. The agent translates it into a single read-only DuckDB query against the snapshot.
Answer
query runs the aggregation and returns typed, structured results your agent can summarise, chart, or drop into a report.
describe_datasetLearn the schema in one callReturns the complete schema: every table with a description, every column with its DuckDB type and meaning, the join keys, and—for enum-like columns (work_model, funding_stage, country, top skills, …)—the actual distinct values present, with counts.
- 23 tables: companies, jobs + funding, skills, comp, locations…
- Real enum vocabularies so the agent filters on exact values
- Row counts + snapshot version for context
- Read-only, idempotent, called once per conversation
queryRun analytics SQLRuns a single read-only DuckDB SQL statement over the full company + job dataset—every nested attribute flattened into typed columns and child tables. Built for aggregation across companies, jobs, funding, skills, comp, and more.
- DuckDB SQL: GROUP BY, percentiles, window functions, joins
- Read-only by design—no writes, no DDL, no file access
- Results capped + flagged so agents aggregate, not page
- Returns typed columns + rows as structured content
Connect your client.
Pick your client below. The first call opens a browser to sign in at enterprise.jobo.world—tokens refresh automatically thereafter.
Add Jobo as a remote MCP server. Claude Code will trigger the OAuth flow on first use.
claude mcp add --transport http jobo https://mcp.jobo.world/mcpThen run `claude mcp list` to confirm the server is registered. The first tool call opens your browser to authorize.
OAuth, the boring (safe) way.
No long-lived API keys to lose. The MCP server speaks OAuth 2.1 with PKCE and Dynamic Client Registration per the MCP spec, so clients self-register and refresh without manual setup.
Discover
Your client fetches /.well-known/oauth-protected-resource on mcp.jobo.world to find the auth server and required scopes.
Authorize
PKCE-protected redirect to enterprise.jobo.world for sign-in. Dynamic Client Registration means no manual app setup.
Call tools
Your agent calls describe_dataset and query with a short-lived JWT. Refresh tokens rotate every 30 days, silently.
Put the whole job market behind your AI.
Add the MCP server to your client in under a minute and start asking. Free to try, scoped to your account, secure by default.