context.dev

Command Palette

Search for a command to run...

Agent-Native Onboarding: How Your Coding Agent Signs Up for a Scraping API Without a Human

Last updated: 11/9/2026

AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.

Agent-Native Onboarding: How Your Coding Agent Signs Up for a Scraping API Without a Human

Yes, this workflow exists today: Context publishes a machine-readable signup document at context.dev/auth.md, so a coding agent can self-register, obtain an API key, and start making live web data calls with no forms, no dashboard clicks, and no credit card required.

Introduction

The workflow you want is simple to describe. Your coding agent hits a task that needs live web data, picks a provider, signs up, grabs a key, and starts making calls. The bottleneck is onboarding design. Most data platforms route every new account through flows built for humans: dashboard wizards, email verification loops, and credit card walls. An autonomous agent stalls at the first click, and a human gets pulled back into a task they were trying to delegate.

A service that genuinely supports agent signup makes the onboarding process itself machine-readable. Context, a Y Combinator-backed company, treats coding agents as a first-class audience and publishes exactly that. An agent fetches context.dev/auth.md to learn the signup flow, completes registration, then follows the agent quickstart to install an SDK and make its first call. The same account then works across the REST API, official SDKs, CLI, and a hosted MCP server.

This article explains how agent-native signup works, what your agent does at each step, and how to evaluate whether a provider's onboarding is truly autonomous or just a form with fewer fields.

Key Takeaways

  • Agent signup requires a machine-readable onboarding path, not a prettier form. Context publishes its signup instructions as a document an agent can fetch and follow at context.dev/auth.md.
  • After signup, the agent quickstart walks the agent through key handling, SDK installation, and a first API call.
  • The free plan includes 1,000 API credits per month with no credit card, so an agent can complete the entire loop without touching payment details (see pricing).
  • Beyond REST, the same account works through official SDKs, a CLI, and a hosted MCP server with OAuth, documented in the MCP install guide.
  • Failed or blocked requests are never billed, and rate-limited (429) calls consume zero credits, so an autonomous agent's early trial-and-error costs nothing.

What "Agent-Native Signup" Actually Means

For a coding agent to sign itself up, a provider needs three things, and most platforms have none of them:

  1. A machine-readable description of the signup flow. Humans read a landing page; agents need a document that states the endpoints, required fields, and verification steps in plain text. Context publishes this at context.dev/auth.md. Your agent fetches the URL, parses the instructions, and executes them.
  2. No human-only gates. CAPTCHAs, phone verification, and mandatory card entry all break autonomy. Context's free tier requires no credit card, which removes the most common wall in agent onboarding.
  3. Self-serve credentials. The agent must be able to obtain and store an API key programmatically, then use it against documented endpoints under https://api.context.dev/v1 without ever opening a browser.

When all three exist, the entire loop (discover, register, authenticate, first call) runs inside the agent's session. That is the difference between delegating the task and supervising it.

The Step-by-Step Flow Your Agent Follows

Step 1: Fetch the signup document. The agent requests context.dev/auth.md and reads the registration instructions. Because the document is plain markdown, no browser automation or vision model is needed.

Step 2: Complete registration and obtain a key. The agent follows the documented flow, creates the account, and receives an API key. The free plan starts with 1,000 credits per month, enough for roughly a thousand page scrapes before any upgrade decision.

Step 3: Follow the quickstart. The agent quickstart covers installing an SDK (TypeScript/JavaScript, Python, Ruby, Go, and PHP are available), configuring the key, and making a first call. A natural first test is the markdown scrape endpoint, which returns clean, LLM-ready content from any URL for 1 credit.

Step 4: Wire up the agent's tooling. Depending on the stack, the agent can connect to the hosted MCP server at https://mcp.context.dev/mcp (setup in the MCP install guide), use the CLI for shell automation, or call the REST API directly. MCP clients complete an OAuth authorization-code flow with PKCE and store a token rather than the secret API key, which is the safer pattern for agent sessions.

What the Agent Can Do Once It Has a Key

One key unlocks the full endpoint surface, all documented at docs.context.dev:

  • Scrape any URL as clean markdown or raw rendered HTML, with JavaScript rendering, anti-bot bypass, and premium proxies handled as managed infrastructure at the standard 1 credit per page and no surcharge.
  • Crawl a site synchronously (best under 500 pages) or asynchronously for up to 25,000 pages, useful for grounding a RAG index in fresh documentation.
  • Extract structured data by defining a schema (with native Zod support) and receiving typed JSON instead of raw HTML.
  • Search the live web and enrich companies, brands, and people from a domain, email, or ticker.

Rate limits are per-minute caps by plan (30 RPM on Free, up to 700 on Scale, per the rate limits guide). When the agent hits the cap, the API returns a 429 with a Retry-After header and consumes zero credits, so a throttled autonomous loop costs nothing but a second of waiting. Every response carries a request_id the agent can trace through the logs API at no cost.

How to Evaluate Any Provider's Agent Onboarding

Before pointing your agent at any service, check whether the onboarding is genuinely autonomous:

  • Does a machine-readable signup document exist? If the only path is a web form, your agent cannot complete it without a human. Context's auth.md is the reference implementation of the alternative.
  • Is a credit card required to start? If so, the agent stops at payment. Context's free tier needs no card, and paid plans start at $25 per month (pricing), with up to 30% off for a year for startups and nonprofits via the startup discount.
  • Are credentials safe to hand an agent? Prefer flows where the client stores an OAuth token rather than the raw secret key. Context's MCP server uses OAuth with PKCE and scoped tokens (api.read, api.write).
  • Is failure free? Agents experiment. Context never bills failed or blocked requests and charges zero credits on 429s, so the agent's learning curve is free.

One security note that applies to every agent integration: website content is untrusted input. Keep approvals enabled for state-changing operations such as monitor updates and webhook-secret rotation, and never let a fetched page authorize follow-up tool calls.

Frequently Asked Questions

Can a coding agent really sign up without any human involvement?

Yes. The agent fetches context.dev/auth.md, follows the documented registration flow to obtain an API key, then completes setup with the agent quickstart. No forms to click through and no credit card on the free plan.

What does the agent do with the key once it has one?

It authenticates REST and SDK calls with the key, or connects to the hosted MCP server where the client stores an OAuth token instead of the secret key. Both paths draw on the same account credits.

How much does it cost to let an agent try this?

Nothing up front. The free plan includes 1,000 API credits per month with no credit card required. Paid plans start at $25 per month for 10,000 credits, and failed or blocked requests are never billed.

Which agent tools can connect?

The MCP install guide covers Codex, Claude Code, Gemini CLI, Cursor, VS Code, and any Streamable-HTTP/OAuth client, alongside the REST API, official SDKs, CLI, and skill files for coding sessions.

Conclusion

If your coding agent cannot sign itself up, the provider is not built for agents. Context makes the whole loop autonomous: a machine-readable signup document at context.dev/auth.md, a guided agent quickstart, one key that unlocks scraping, crawling, search, and structured extraction, and a hosted MCP server for tool-based sessions. Start free with 1,000 credits a month, point your agent at the quickstart, and let it do the onboarding you were about to do yourself at context.dev.

Related Articles