What's shipped.

An interactive playground at anakin.io/demo lets developers test scrape, map, crawl, search, and AI search against any URL before creating an account. Five endpoint modes, two free anonymous calls, then a contextual signup prompt. No credit card required.

Five new official SDKs published across major registries:

• Rust — crates.io/crates/anakin-sdk cargo add anakin-sdk
• PHP — packagist.org/packages/anakin/sdk composer require anakin/sdk
• Ruby — rubygems.org/gems/anakin-sdk gem install anakin-sdk
• Elixir — hex.pm/packages/anakin {:anakin, "~> 0.1"}
• Java — Maven Central io.github.anakin-inc:anakin-sdk:0.1.0

All five mirror the same surface as the existing Python, Node.js, and Go SDKs: scrape, map, crawl, search, agentic_search, wire, browser-session CRUD, and a typed error hierarchy.

Dedicated reference pages added for Go, Rust, PHP, Ruby, Elixir, Java, and .NET (coming soon). Each covers install, quickstart, method reference, configuration options, and the full error hierarchy. The SDKs landing page now lists all nine languages with install commands.

The official Go module is now live on the Go module proxy and pkg.go.dev.

go get github.com/Anakin-Inc/anakin-go

Uses functional options (WithAPIKey, WithTimeout, etc.), zero runtime dependencies (pure stdlib HTTP), and 23 tests. Requires Go 1.21+.

Previously, npx -y @anakin-io/mcp init --all exited with an error if ANAKIN_API_KEY wasn't set — forcing a manual export step. The --all flag controls per-client confirmations, not the key prompt. Now the key is always requested interactively if the env var is absent, regardless of mode.

Four packages went live simultaneously:

• @anakin-io/sdk v0.1.0 — Node.js / TypeScript SDK on npm
• @anakin-io/mcp v0.1.0 — Model Context Protocol server on npm (auto-configures Claude Desktop, Claude Code, Cursor, Cline, Continue, Zed, Windsurf, VS Code)
• anakin-sdk v0.1.0 — Python SDK on PyPI (import name: anakin)
• anakin-cli v0.2.0 — CLI for terminal-driven scraping on PyPI

All packages are Apache 2.0 licensed and open source under github.com/Anakin-Inc.

New endpoints to discover and search the Wire (Holocron) action catalog:

• GET /v1/holocron/catalog — list available action catalogs
• GET /v1/holocron/catalog/{id} — get a specific catalog
• POST /v1/holocron/search — search actions by keyword

Enables programmatic discovery of pre-built Wire actions without needing the dashboard.

Browser Connect (/v1/browser-connect) now supports loading saved sessions via ?session_id or ?session_name. Cookies, localStorage, and proxy assignment are injected into the CDP context before the WebSocket upgrade — eliminating re-login flows for authenticated scraping targets.

anakin-cli v0.2.0 ships:

• New research command powered by agentic search — ask a question, get a synthesised answer with sources
• Markdown output now renders in-terminal with colour coding
• --format json flag on all commands for piping to jq
• --country flag on scrape and crawl for geo-targeted results

POST /v1/agentic-search now accepts an optional schema field (JSON Schema object). When provided, the answer is returned as structured JSON matching the schema rather than free-form text. Useful for extracting typed data from multi-source research tasks.

Default rate limits increased across the board:

Enterprise plans can request higher limits via the dashboard.

POST /v1/map gains two new parameters:

• include_subdomains (bool) — discover links across all subdomains of the root domain
• search (string) — filter returned links to those matching a keyword (server-side, no extra API calls)

Both are optional and backward-compatible.

Pass "screenshot": true in the scrape request body to capture a full-page PNG screenshot of the scraped page. Returned as a base64-encoded string in the screenshot field of the response. Automatically handled by the browser worker when useBrowser is also true.

The browser worker now runs Camoufox (Firefox-based) instead of Chromium. Key improvements:

• Better fingerprint masking — canvas, WebGL, audio, and font fingerprints are randomised
• WebRTC IP leak prevention enabled by default
• Automatic CAPTCHA solving via CapSolver extension (reCAPTCHA v2/v3, hCaptcha, AWS WAF)
• Watchdog auto-restarts with exponential backoff on browser crash

POST /v1/crawl now accepts:

• max_depth (int, default 2) — how many link-hops from the start URL to follow
• max_pages (int, default 100) — hard cap on total pages crawled

Both are enforced server-side. The job completes when either limit is hit or the site is exhausted.

Previously, credits were checked and reserved upfront but deducted at the end. A race condition meant some failed jobs still consumed credits. The deduction path is now strictly gated on status = completed. Jobs that timeout, fail, or are cancelled consume zero credits.

A new internal service uses Thompson Sampling (Bayesian multi-armed bandit) to route scrape requests to the best-performing proxy for each domain+country pair. Proxy scores decay over 24h toward a neutral prior. The result: lower block rates on difficult targets without manual proxy management.

API keys are now stored AES-256-GCM encrypted at rest. The SHA-256 hash is used for fast lookup; the plaintext key is only shown once at creation time. Existing keys were migrated transparently — no action required from existing users.

POST /v1/url-scraper/batch submits up to 10 URLs in a single API call. Each URL is processed as an independent job; the batch endpoint returns a list of job IDs for polling. Useful for scraping product listings, search results, or other sets of known URLs.

The dashboard now shows a 30-day credit usage chart and a per-request breakdown table (endpoint, URL domain, credits used, timestamp). Filterable by endpoint type. Useful for understanding spend distribution before upgrading plans.