Skip to main content
The context-dev CLI wraps every Context.dev endpoint in a single binary. Anything that can run a shell command — you, a CI job, or a coding agent like Claude Code or Codex — can resolve brands, scrape pages to Markdown, take screenshots, and classify industries without writing any SDK code.

Install

Prebuilt binaries for macOS, Linux, and Windows ship with every GitHub release. If you have Go 1.22+ installed, go install works too.
terminal
The archive also bundles shell completions (bash, zsh, fish) and man pages if you want them.
Confirm it installed:
terminal

Set your API key

Sign up at context.dev/signup (free tier: 500 one-time credits, no card) and copy your key from the dashboard — it starts with ctxt_secret_. The CLI reads the key from the CONTEXT_DEV_API_KEY environment variable. Every command also accepts an explicit --api-key flag, which takes precedence over the environment.

Temporary (current terminal only)

export sets the key for the shell you’re in; it disappears when you close the terminal:
terminal
For a single command, prefix the variable or pass the flag — neither touches your shell state:
terminal

Permanent (every new terminal)

Append the export to your shell profile so the key is set in every session:
terminal
On Windows, setx persists it for all future terminals:
terminal
Never commit the key. Keep it in your shell profile, an .envrc, or a secret manager — and never bake it into scripts that land in a repo.

Usage

Commands follow a resource-based structure:
terminal
ResourceWhat it covers
brandBrand profiles by domain, name, email, ticker, or ISIN; transaction identification
webScraping (Markdown, HTML, images, sitemap, crawl), search, screenshots, styleguide, fonts, competitors, JSON Schema extraction
aiProduct extraction and AI data queries
industryNAICS and SIC classification
utilityZero-credit cache prefetching
--help works at every level: context-dev --help lists the resources, context-dev web --help lists its commands, and context-dev web web-scrape-md --help documents every flag. Three global flags do most of the output work:
terminal
--format accepts auto, explore, json, jsonl, pretty, raw, and yaml; auto (the default) picks based on whether output is going to a terminal or a pipe. --transform uses GJSON syntax to reshape responses before they’re printed.

Examples

One example per endpoint, grouped the same way as the API Reference. Credit costs are noted inline; the full breakdown is on the pricing page.

Brand Intelligence

terminal
See Get brand data and Enrich transaction codes for what comes back.

Web Scraping

terminal
See Scrape websites to Markdown and Take webpage screenshots.

Web Extraction

terminal
See Extract a design system and Extract structured data.

Product Extraction & AI Queries

terminal
See Extract products from websites.

Industry Classification

terminal
See the classification overview for how the three systems differ.

Prefetch

terminal
See Prefetching for when this pays off.

Give it to your agent

Any agent with shell access can use the CLI as-is — no MCP server or tool config required. Make sure CONTEXT_DEV_API_KEY is exported in the shell your agent inherits, then tell it the CLI exists in your CLAUDE.md or AGENTS.md:
CLAUDE.md / AGENTS.md
When the agent pipes a command, --format auto emits plain JSON, so the output is parseable without extra flags. For trimming large responses before they hit the agent’s context window, --transform is the tool: --transform 'brand.colors' returns just the colors instead of the full profile.

Next steps

Agent Quickstart

The full agent setup: copyable system prompt, SKILL.md, and MCP, in one page.

Install MCP

Prefer tool calls over shell commands? Connect the MCP server instead.

SKILL.md

Teach your agent when to reach for each endpoint.

API Reference

The underlying endpoints every CLI command calls.