Everything Claude Code

```json
{
  "service_type": "platform",
  "base_url": "https://ecc.tools",
  "auth_method": "unknown",
  "auth_config": {
    "notes": "No public documentation was provided regarding authentication. Likely a browser-based dashboard or CLI tool. Auth may be required for hosted features."
  },
  "endpoints": [],
  "pricing_model": {
    "type": "unknown",
    "details": {
      "notes": "No pricing page or model was included in the available content. ECC positions itself as a developer tooling layer for agent harnesses; likely freemium, open-source, or subscription-based, but unconfirmed."
    }
  },
  "rate_limits": {
    "notes": "Unknown. No API surface identified."
  },
  "capabilities": [
    "Agent harness performance optimization",
    "Skill management for AI coding agents",
    "Instinct/primitives library for agent behavior tuning",
    "Persistent memory layer for agents",
    "Security tooling for agent runs",
    "Research-first development workflows",
    "Multi-agent support (Claude Code, Codex, Opencode, Cursor)",
    "Developer-focused search and discovery"
  ],
  "raw_analysis": "Everything Claude Code (ECC) at https://ecc.tools is a developer-tooling platform that layers performance optimization onto popular AI coding agent harnesses — Claude Code, OpenAI Codex, Opencode, and Cursor. Its stated feature areas are: Skills (reusable capability packs), Instincts (behavioral heuristics/primitives), Memory (persistent context across sessions), Security (guardrails for agent execution), and a research-first development methodology.\n\nPositioning: This is infrastructure/dev-experience tooling for engineers running agentic coding workflows. It sits one layer above the underlying LLM providers and agent CLIs — it does not appear to be an LLM provider itself. Its value proposition is standardization and performance uplift across heterogeneous agent harnesses, which is a real pain point as teams adopt multiple agents.\n\nMaturity: The category ('agent harness optimization') is very young (2024-2025 vintage). The tagline markets it as the category-defining 'system' for this, but that is self-asserted. No funding, changelog, docs, or version history was supplied, so real maturity is hard to verify. The explicit multi-harness support (Claude Code, Codex, Opencode, Cursor) is a meaningful signal that it's not vaporware — those integrations are non-trivial.\n\nIntegrations: Explicitly targets Claude Code, Codex (OpenAI), Opencode, and Cursor. 'And beyond' suggests extensibility, likely via a plug-in or config-based interface. No MCP server, REST API, or SDK is publicly documented in the provided material.\n\nAPI surface: None discovered. No OpenAPI spec, no public REST endpoints, no documentation link. Interaction is likely through a CLI, npm package, or browser extension that hooks into each agent harness — not via HTTP endpoints we can call.\n\nAuth: Unknown. If there is a hosted dashboard (ecc.tools is a web domain), it likely uses email magic-link or OAuth (GitHub/Google), common for dev tools. If it's purely a local CLI/npm package, auth may be a license key or nothing at all.\n\nPricing: Not disclosed. Competitors in this space (Cursor extensions, agent frameworks like LangChain/Jitera, and tooling from Anthropic/OpenAI) tend toward freemium or OSS-core models. Best guess: freemium with paid team/security tiers, but this is inference, not observation.\n\nRisks / caveats for integration: (1) No public API — automation against ECC likely requires wrapping their CLI or config files. (2) Dependency on third-party agent harnesses means churn risk if Claude Code, Cursor, or Codex change their plugin surfaces. (3) 'Security' claims need independent verification before trusting with credentials or sensitive repos. (4) Single-page marketing site with no docs shown; recommend verifying docs, license, and security posture before adoption.\n\nRecommendation: Treat as a promising but under-documented platform. For a Prowl integration, we should first probe /docs, /pricing, and any GitHub org for an SDK or MCP endpoint before assuming REST access. If none exists, integration would need to be at the CLI/config layer, which is higher-effort and more brittle than an API."
}
```

2/3 tests passed

```json
{
  "overall": 62,
  "dimensions": {
    "token_efficiency": 7.0,
    "first_try_success": 6.0,
    "response_parseability": 5.0,
    "error_clarity": 5.0,
    "doc_quality": 6.0,
    "auth_simplicity": 6.0,
    "latency": 10.0,
    "consistency": 8.0
  },
  "pricing_normalized": {
    "model": "unknown",
    "notes": "No pricing page surfaced in available content. Likely freemium/open-source/subscription for agent harness tooling, but unconfirmed — a blocker for agent recommendations requiring cost transparency."
  },
  "issues": [
    "No llms.txt (404) — agents relying on machine-readable docs can't fetch a canonical LLM-oriented summary.",
    "Pricing model undisclosed: no pricing page detected, which is a hard gating factor for recommendations.",
    "Positioning phrase 'The #1 agentic engineering toolkit' is marketing-heavy and vague; unclear what ECC concretely does vs. adjacent tools.",
    "No structured API/export presented in the available content — parseability is inferred from tooling nature, not confirmed via endpoints.",
    "Capability list is broad (memory, security, skills, multi-agent, research) — scope is diffuse, harder to convey concisely to users.",
    "No visible auth method details (SSO, magic link, Notlogin) — onboarding friction unknown."
  ],
  "recommendations": [
    "Publish /llms.txt with a terse capability + endpoint + auth summary so agents can ingest without scraping HTML.",
    "Add a public /pricing page (or state open-source license) to unblock cost-driven recommendations.",
    "Sharpen the one-line value prop: specify the concrete primitive (e.g., 'instincts + memory layer for Claude Code/Codex/Cursor') rather than '#1 toolkit'.",
    "Expose a machine-readable manifest (OpenAPI or JSON capabilities endpoint) to lift response_parseability.",
    "Document supported agent host matrix (Claude Code, Codex, Opencode, Cursor) with version constraints for first-try success.",
    "Adopt Notlogin-style verified agent onboarding or magic-link auth to push auth_simplicity toward 10.",
    "Publish a status page / SLA note to substantiate reliability claims beyond single 200-response uptime checks."
  ]
}
```