feat!: rename MCP tools to match v2 docs (v3.0.0, BREAKING)

[Vikrant-Khedkar]

Summary

Aligns the MCP tool surface to the canonical v2 API vocabulary at docs.scrapegraphai.com/services/*. The MCP currently exposes legacy names (smartscraper, searchscraper, smartcrawler_initiate, …) while the docs and SDKs use extract, search, crawl_start, etc. This PR fixes that drift.

Hard rename, no aliases. Bumped to v3.0.0.

Renames

v2 (old)	v3 (new)	Endpoint
smartscraper	extract	POST /extract
searchscraper	search	POST /search
smartcrawler_initiate	crawl_start	POST /crawl
smartcrawler_fetch_results	crawl_get_status	GET /crawl/:id
sgai_history	history	GET /history
generate_schema	schema	POST /schema
markdownify	removed	use scrape with output_format="markdown"

Unchanged: scrape, crawl_stop, crawl_resume, credits, monitor_* (already match docs).

Tool count: 18 → 17.

What changed

• src/scrapegraph_mcp/server.py — renamed 6 @mcp.tool functions, deleted markdownify tool. Also removed dead code on ScapeGraphClient: the markdownify() HTTP method and two backwards-compat shims (extract/search) that would have recursed infinitely after the rename.
• pyproject.toml, server.json — version → 3.0.0.
• README.md — added v2→v3 migration section, updated all tool tables and prose.
• .agent/README.md, .agent/system/project_architecture.md, .agent/system/mcp_protocol.md — swept all references.
• uv.lock — picked up the version bump.

Test plan

End-to-end verified locally via HTTP transport with X-API-Key:

• tools/list returns exactly the 17 canonical names
• credits returns live API data
• extract (was smartscraper) on https://scrapegraphai.com returns structured JSON
• search returns results
• scrape returns markdown
• history returns recent request log
• Old names (smartscraper, markdownify) return Unknown tool: <name> (proves no silent fallback)
• ruff check src/ passes
• No stale old-name refs anywhere outside the README migration table

Out of scope (follow-ups)

• Mintlify docs at docs.scrapegraphai.com/services/mcp-server/introduction live in a separate repo — needs a parallel docs PR before v3 ships publicly.
• After merge: tag v3.0.0, publish to PyPI, Smithery picks up the new package on redeploy.

🤖 Generated with Claude Code