Official Apify plugin for Claude Code. It adds the Apify MCP server, one apify routing agent, and five bundled skills for the main Apify workflows: using existing Actors, building and deploying your own Actors, actorizing existing projects, generating Actor output schemas, and integrating Apify into existing applications.
Apify is the largest marketplace of tools for AI: ready-made Actors you can run, or build your own. Find your Actor at Apify Store.
| Component | Name | Purpose |
|---|---|---|
| Agent (entry point) | apify |
Routes every Apify request to the right skill or transport path. This is the one you should invoke. |
| MCP server | apify (https://mcp.apify.com/) |
Lets the agent search the Apify Store, fetch Actor details, run Actors, and read the Apify docs. |
| Skill | apify-actor-development |
Create, debug, and deploy a brand new Apify Actor from scratch. |
| Skill | apify-actorization |
Convert an existing JS/TS, Python, or CLI project into an Apify Actor. |
| Skill | apify-generate-output-schema |
Generate dataset_schema.json / output_schema.json / key_value_store_schema.json for an existing Actor. |
| Skill | apify-sdk-integration |
Add Apify Actor execution to an existing application using the apify-client package. |
| Skill | apify-ultimate-scraper |
CLI-driven data extraction workflow for selecting, configuring, and running pre-built Actors across 15+ platforms. |
Add the Apify marketplace and install the plugin:
/plugin marketplace add apify/apify-claude-code-plugin
/plugin install apify@apifyValidate the installation:
/plugin validate .- Claude Code CLI installed and authenticated.
The plugin uses three setup paths depending on what you're doing. The apify agent will guide you through the right one, but here is the high-level map.
Uses OAuth. The first time the agent calls a tool that needs auth (for example run-actor or get-dataset-items), Claude Code opens console.apify.com in your browser and asks you to sign in. Read-only tools such as search-actors, fetch-actor-details, search-apify-docs, and fetch-apify-docs work without auth.
These skills expect the local apify CLI to be available. Install it first:
npm install -g apify-cliFor interactive use, authenticate with:
apify loginIn headless or CI environments, export an APIFY_TOKEN instead; the CLI can read it automatically:
export APIFY_TOKEN="apify_api_xxxxxxxxxxxx"Generate a token at console.apify.com/settings/integrations. Don't have an account? Sign up free — no credit card required.
Uses an APIFY_TOKEN environment variable with the apify-client package or the REST API:
export APIFY_TOKEN="apify_api_xxxxxxxxxxxx"The MCP OAuth flow needs a browser. If you're running Claude Code over SSH or in any environment without a browser, you have these options:
- Authenticate locally first. Run Claude Code once on your laptop so the OAuth refresh token is stored, then reconnect remotely.
- Use the CLI-based skills.
apify-actor-development,apify-actorization, andapify-ultimate-scrapercan work in headless environments when theapifyCLI is installed andAPIFY_TOKENis exported. - Use the SDK integration skill.
apify-sdk-integrationusesapify-clientand only needsAPIFY_TOKEN.
Start a Claude Code session and describe what you need. The apify agent reads the request, chooses between MCP, CLI, or SDK-based workflows, and dispatches to the right skill or tool.
find me 5 well-rated coffee shops in Seattle and export to CSV
build me an Actor that scrapes a sitemap and stores titles
add Apify to this Next.js app so I can run a scraper from /api/scrape
generate output schemas for the Actor in this folder
The apify MCP server is configured in .mcp.json and the agent uses it for Route 1 tasks. It exposes:
search-actors— search the Apify Store by keyword (no auth)fetch-actor-details— Actor specs, input schema, pricing (no auth)run-actor— execute an Actor and return results (OAuth)get-dataset-items— retrieve dataset rows from a previous run (OAuth)search-apify-docs/fetch-apify-docs— Apify documentation lookup
This plugin export does not include standalone helper scripts. Instead, the bundled skills ship with markdown references that the agent uses while working:
skills/apify-actor-development/references/— Actor config, schemas, logging, standby mode, and README guidanceskills/apify-actorization/references/— JS/TS, Python, and CLI actorization guides plus schema/output notesskills/apify-ultimate-scraper/references/— Actor index, gotchas, and workflow playbooks for common scraping use cases
The executable requirements come from the skills themselves: Route 1 uses the MCP server, Route 2 relies on the local apify CLI, and Route 3 uses the apify-client SDK.
The plugin ships only text files (Markdown, JSON, and the license) and an icon image. It has no hooks, scripts, or executables, so it has no code of its own that reads, stores, or sends your Apify credentials. Everything it does goes through one of the three setup paths:
| Path | What leaves your machine | Where it goes | Which credential is used |
|---|---|---|---|
| Path 1 — MCP server | Your Actor searches, Actor inputs, and requests for run results and docs | https://mcp.apify.com |
OAuth sign-in with your Apify account. Claude stores the OAuth token, not the plugin. |
| Path 2 — Apify CLI | Commands Claude runs with your local apify CLI, for example Actor runs and apify push, which uploads your Actor's source code |
Apify API | The CLI's own login from apify login, or APIFY_TOKEN if you set it for headless use |
| Path 3 — SDK integration | Nothing at setup. Claude writes code into your application. | Apify API, when you run your application | Your application reads APIFY_TOKEN from its own environment |
Also good to know:
- Your Apify account does the work. Actors run on the Apify platform under your account and use its credits, including paid Actors. Run results are stored in your account's datasets and key-value stores, and the results Claude asks for are returned into the conversation.
- Usage attribution. The MCP server URL includes
?client=claude+code+plugin, and Actor runs started through the CLI pass--user-agent apify-claude-code-plugin/…, so Apify can see that a request came from this plugin. - Public docs. To answer questions, Claude may fetch public pages such as
https://docs.apify.com/llms.txtand Apify Store pages. No credentials are sent with these requests.
Apify processes this data under its privacy policy.
OAuth browser never opens / hangs. See the "Working in headless / SSH environments" section above and switch to a CLI- or SDK-based path if needed.
apify CLI not found. Install it with npm install -g apify-cli before using apify-actor-development, apify-actorization, or apify-ultimate-scraper.
APIFY_TOKEN not found. Export APIFY_TOKEN in your shell before starting Claude Code when using headless CLI auth or the apify-sdk-integration skill.
The wrong skill keeps getting picked. The five bundled skills are marked user-invocable: false and are meant to be routed through the apify agent. If routing looks wrong, describe the goal more explicitly: use existing Actors, build an Actor, actorize a project, generate output schemas, or integrate Apify into an app.
apify vs apify-client — these are two different npm packages. The apify package is the SDK for building Actors (used inside an Actor's code, on the Apify platform). The apify-client package is the API client for calling Actors from your own application. The agent picks the right one for you; if you're installing manually, double-check.
- Apify Console — console.apify.com
- Apify Store — apify.com/store
- Docs (LLM-friendly) — docs.apify.com/llms.txt
- Docs (full) — docs.apify.com/llms-full.txt
- Source repo — github.com/apify/apify-claude-code-plugin
- Issues / feedback — open an issue on the source repo, or email support@apify.com
Apache-2.0. See LICENSE.