v2.0.0 — Strategic Briefing replaces Analysis

⚠️ Breaking change

The get_action_plan tool has been removed — the underlying /v1/projects/{projectId}/analysis/action-plan endpoint no longer exists in the CompetLab API. It's replaced by the new get_strategic_briefing tool.

✨ New tool — get_strategic_briefing

The synthesized, prioritized competitive read across all dimensions: what changed, what it means, and what to do about it.

• Defaults to the executive hub digest (headline + per-dimension verdicts naming the deeper section to open) — answers most questions in one call.
• Drill deeper with sections: actions, competitors, any of the 13 deep-<dimension> slots, or all.
• includeCharts (default false) adds full chart series.
• Branch on meta.availability: ready · ready-refreshing · preparing · none. coverage + dimensionHealth are always returned.

Example prompt:

"Show me the strategic briefing — what should I fix first?"

🧹 Also in this release

• Corrected all tool-count docs to the accurate 33 tools / 30 read-only (several had drifted to 32/29).
• Tool count unchanged — still 33 tools, zero new dependencies.

Connect

Hosted users (https://mcp.competlab.com/mcp) need do nothing. Self-hosters: git pull && npm run build.

Full diff: v1.2.0...v2.0.0

v1.2.0 — fetch_url + API 2026-05-22 sync

Description

Mirrors the @competlab/sdk@1.2.0 release. Brings the MCP surface in line with the 2026-05-22 API ship.

✨ New tool: fetch_url

Fetch any URL with automatic JS-rendering and common bot-protection handling. Returns body, headers, and cleanStats. Optional cleanHtml strips HTML noise while preserving text content — significant token-cost win for LLM consumption (typical: ~90% reduction on content-heavy pages).

• Rate limit: 60 req/min per API key (tighter than the 1000/min default for other free tools)
• Honest degradation: headersAvailable: false is surfaced when targets use advanced behavioral fingerprinting
• Pass-through validation: out-of-range knobs, IPv4/IPv6/localhost hosts, and the no-output combo (bodyNeeded: false, headersNeeded: false) return precise invalid_parameters errors

📡 New degradation discriminators (pass-through)

The MCP server stays a thin pass-through over the API. The 2026-05-22 response-shape additions flow through automatically — no client changes needed unless you want to handle the degraded case explicitly:

• signalsAvailable on get_trust_signals_scan and per-competitor security headers
• partialDetection on get_tech_stack_scan (HTML-based detection still fires; only header-based hosting/CDN signals are affected)
• securitySignalsAvailable + nullable securityGrade / securityScore on get_tech_trust_dashboard
• details.reason: 'site_uses_behavioral_protection' on agent-adoption link-headers / cache-header-hygiene checks (scored: false, weight: 0 — no score impact)
• New BEHAVIORAL_PROTECTION_ADOPTED alert type + previousSecurityProfile context field on list_alerts
• Counter rename: meta.counters.n8nCalls → meta.counters.heavyFetchCalls on get_agent_adoption_scan

Tool descriptions for the 5 affected tools now include 1-line advisory notes so agents handle the degraded case cleanly without re-reading the spec.

📊 Counts

• Tool count: 32 → 33
• Free tools: 8 → 9
• Read-only: 29 → 30

✅ Verified

Smoke-tested against prod:

• competlab.com (healthy) → headersAvailable: true, full body + headers, cleanStats showing 92.95% reduction
• g2.com (bot-protected) → headersAvailable: false, body populated, headers correctly absent
• Negative test (bodyNeeded: false, headersNeeded: false) → returns the locked 400 invalid_parameters response verbatim

🔗 Companion release

• SDK: @competlab/sdk@1.2.0 (already published 2026-05-22)

v1.1.1 — Free Tools (Wave 2)

8 new MCP tools — free tools that work on any public domain, no project setup required. Server catalog grows from 24 → 32 tools. Mirrors monorepo PRs #179 + #180.

• check_sitemap, check_ai_crawlers — sync live checks
• start_/get_tech_stack_scan — async tech-stack detection (117 rules)
• start_/get_trust_signals_scan — async trust-signals analysis (34 signals)
• start_/get_agent_adoption_scan — async Agent-Adoption Check (25 checks)
• 29 read-only + 3 async-start tools (3 start_*_scan tools create scan records)
• Industry parameter accepts forward-compatible string values
• Migrated to server.registerTool() (non-deprecated SDK API)
• Live-tested end-to-end against api.competlab.com

v1.0.0 — Initial Release

CompetLab MCP Server guide repo — documentation, configs, and metadata for connecting AI agents to CompetLab competitive intelligence.

• 24 MCP tools across 5 dimensions
• Streamable HTTP transport
• Config examples for Claude, Cursor, VS Code, Windsurf, Cline
• Published on Smithery, Glama, mcp.so, and more