AI agents & API

The two-minute MCP setup.

Connector or API key - every client works.

One copied prompt that configures any agent.

Endpoints, auth, and rate limits.

npx argorant - search, verify and export from the terminal.

What they are, and the fair-use limits.

What an agent can do, and what it can never do silently.

The full connector walkthrough, with what to check when it does not appear.

One command, plus the CLI pairing that makes terminal work fast.

The MCP path, and which ChatGPT plans expose it.

The API key path that works on every plan.

MCP where available, REST or CLI everywhere else.

Add the MCP server, or paste the setup prompt and let Cursor configure itself.

Connector where xAI enables MCP, bearer key everywhere else.

xAI's terminal coding agent, wired to your data layer.

Always-on teammates with a real data layer and a hard budget.

Custom runtimes: which surface to pick and how to shape the loop.

What that sign-in pop-up is, why PKCE, and why some clients redirect to cursor://

What each instruction in the copied prompt does, and why it is written that way.

Identity, scopes and daily quota. The call every agent should make first.

Free market counts. The endpoint you should call ten times before spending once.

Free redacted previews. See the actual people before you pay for them.

The paid endpoint. One credit per contact, full details returned inline.

Create a verified CSV export job and get a job id back.

Polling a job, fetching the CSV, and handling exports above 50,000 rows.

Company counts, company search, and who works at a given domain.

Verify your own addresses, one at a time or in batches of 500.

Save a filtered list, then check its size later. Both free.

One means the key is wrong, the other means the key is not allowed.

Nothing is broken. The account just cannot pay for what you asked.

Back off, then resume. Nothing is lost.

What npx argorant is, how it authenticates, and the one rule to internalize.

Three ways to authenticate, and which one wins.

The two free commands you should wear out before spending anything.

Qualify a target account in one command, for free.

The command that spends credits inline. Read this before scripting it.

Verified CSVs from the terminal, including exports above 50,000 rows.

Check addresses you already own, one at a time or by the thousand.

Save a segment as a filter set, check its size any time. Both free.

The complete flag surface, plus the two flags that behave differently than you expect.

Six codes that let an agent branch without parsing text.

What this command group is and why your key probably cannot use it.

One key per agent, rotated on a schedule, revoked without drama.

Copy-ready rules for your agent's system prompt, plus the limits enforced server-side.