Apify for Kilo

@apify/kilocode-plugin gives Kilo Code agents one native tool — apify — that can run any of Apify's 20,000+ web-scraping and automation Actors and bring the results back into the conversation. Scrape Google Maps, Instagram, Amazon, TikTok, search the web, crawl a site, or invoke any Actor by slug — without a tool-per-site explosion.

Community integration. "Apify for Kilo" is built and maintained by Apify and is not officially affiliated with Kilo.

Install

Add a single line to your kilo.jsonc:

{
  "plugin": [
    ["@apify/kilocode-plugin", { "token": "{env:APIFY_TOKEN}" }]
  ]
}

Then set your token (get one at console.apify.com/account/integrations):

export APIFY_TOKEN="apify_api_..."

Or install via the CLI, which patches your config for you:

kilo plugin @apify/kilocode-plugin

Token sources

The plugin resolves the token in this order:

1. Kilo auth login — the plugin registers an "Apify API Token" auth method, so you can store the key through Kilo's native auth flow:

   kilo auth login --provider apify   # paste your apify_api_… token

   This alone is enough — login enables the tool on the next launch, with no config or env var needed.

2. Config — { "token": "..." } (or { "apiKey": "..." }) in the kilo.jsonc plugin entry.

3. Environment — APIFY_API_KEY or APIFY_TOKEN.

If no token is found, the plugin registers no tool (the agent never sees a tool that would fail). After running kilo auth login, restart Kilo so the plugin re-reads the credential.

The apify tool

One tool, three actions. The agent composes them:

| Action | What it does | |------------|--------------| | discover | Search the Apify Store (query), or fetch an Actor's input schema + README (actorId). | | start | Launch an Actor run with a JSON input; returns run references immediately (non-blocking). | | collect | Poll the runs; return finished datasets, mark the rest pending, surface failures. |

Typical flow: discover (search) → discover (schema) → start → collect (repeat collect while allDone is false).

Example (what the agent does under the hood)

// 1. Learn the Actor's inputs
{ "action": "discover", "actorId": "compass~crawler-google-places" }

// 2. Launch a run (batch many targets into ONE run when possible)
{ "action": "start",
  "actorId": "compass~crawler-google-places",
  "input": { "searchStringsArray": ["coffee shop Prague"], "maxCrawledPlacesPerSearch": 3 },
  "label": "prague-coffee" }

// 3. Collect — repeat until allDone: true
{ "action": "collect",
  "runs": [{ "runId": "…", "actorId": "compass~crawler-google-places", "datasetId": "…", "label": "prague-coffee" }] }

Slugs use a tilde: username~actor-name (e.g. apify~google-search-scraper), never a slash.

Safety

• Prompt-injection defense. Scraped data is wrapped in <<<EXTERNAL_UNTRUSTED_CONTENT>>> markers (with the boundary markers sanitized out of the body) so a host that respects them won't treat scraped text as instructions.
• Result-size cap. Dataset output is truncated to 50,000 characters to protect the context window.
• SSRF guard. baseUrl must start with https://api.apify.com.
• Secret hygiene. Tokens are newline-stripped before use and never echoed in tool output.

Configuration

| Field | Purpose | |----------------|---------| | token / apiKey | Apify API token. Falls back to APIFY_API_KEY / APIFY_TOKEN. | | enabled | Force on/off. When unset, the plugin enables itself iff a token is present. | | baseUrl | Override the Apify API origin. Must start with https://api.apify.com. | | maxResults | Default result cap (forward-looking; currently informational). | | enabledTools | Subset of tools to register (only apify exists today). |

Development

npm install
npm run build       # tsc → dist/
npm run typecheck
npm test            # vitest, mocked apify-client

See CLAUDE.md for architecture notes and local end-to-end testing against a Kilo checkout.

License

Apache-2.0 © Apify