Plandex

```json
{
  "service_type": "platform",
  "base_url": "https://plandex.ai",
  "auth_method": "none",
  "auth_config": {},
  "endpoints": [],
  "pricing_model": {
    "type": "freemium",
    "details": {
      "open_source": true,
      "notes": "Core project is open source (self-hostable, bring your own LLM API keys). A hosted cloud offering exists with a freemium model where users pay for usage/credits."
    }
  },
  "rate_limits": {
    "notes": "Self-hosted: constrained by the underlying LLM provider's rate limits. Cloud-hosted: governed by Plandex Cloud quota/credit tiers (not publicly documented as fixed numeric limits)."
  },
  "capabilities": [
    "AI coding agent for large-scale, multi-file projects",
    "Terminal/CLI-based agent (plandex CLI) rather than REST API",
    "Sandboxed execution with review/approval workflow",
    "Context management for large codebases (context loader/packer)",
    "Multi-model / multi-provider LLM support (OpenAI, Anthropic, Google, open-source models, local via Ollama)",
    "Git-aware workflows and versioned AI plans",
    "Self-hostable (open source) and cloud-hosted option",
    "Bring-your-own API key support"
  ],
  "raw_analysis": "Plandex (plandex.ai) is an open-source AI coding agent aimed at developers working on large, real-world software projects. Rather than a SaaS platform exposing a public REST API, it primarily presents as a terminal/CLI tool (the `plandex` binary) that coordinates an AI agent to plan and execute multi-step coding tasks across many files. Because it is distributed as an installable CLI, it does not expose a public HTTP API surface that Prowl can integrate with in the conventional sense; integration is done by installing and running the tool locally or via its hosted cloud service.\n\nWho it's for: individual developers and teams who want an agentic coding assistant that can handle larger codebases than single-file copilots, with a focus on controllable, reviewable changes rather than blind automation.\n\nMaturity: The project is open source and actively developed, with a real user base and a hosted cloud tier. It is relatively young compared to mainstream IDE copilots, so the ecosystem (IDE plugins, third-party integrations) is thinner, though it supports a broad set of LLM providers and a bring-your-own-key model.\n\nIntegrations: Primary integration points are (1) the underlying LLM providers (OpenAI, Anthropic, Google, and local models via Ollama), (2) the local filesystem/git repositories it operates on, and (3) its cloud service for those who don't want to self-host. There is no documented public REST API, webhook system, or OAuth flow, so programmatic integration should be assumed to occur through the CLI and its on-disk project state, not over HTTP endpoints.\n\nPricing: Core tool is open source and free to self-run (costs shift to LLM provider usage). The managed cloud offering is freemium/credit-based. Exact rate limits and pricing tiers are not publicly documented in fixed numeric terms and would need to be verified from the Plandex cloud pricing page.\n\nRecommendation for Prowl: Treat Plandex as a CLI/self-hosted 'platform' with no public REST API. Integration, if any, would be via shelling out to the CLI or embedding it, not via declared HTTP endpoints."
}
```

0/3 tests passed

{
  "overall": 62,
  "dimensions": {
    "token_efficiency": 8.0,
    "first_try_success": 5.0,
    "response_parseability": 6.0,
    "error_clarity": 6.5,
    "doc_quality": 7.0,
    "auth_simplicity": 6.0,
    "latency": 2.0,
    "consistency": 4.0
  },
  "pricing_normalized": {
    "model": "freemium",
    "open_source": true,
    "self_hostable": true,
    "byo_api_key": true,
    "notes": "Open-source core with BYO LLM keys; hosted cloud tier with usage-based credits."
  },
  "issues": [
    "All three live checks (uptime, robots.txt, llms.txt) failed with DNS resolution errors, indicating the tested domain could not be reached — agents cannot verify availability or crawl policy.",
    "CLI-first, agentic architecture with no REST API surface makes it hard for AI agents to programmatically recommend or integrate it on a user's behalf.",
    "No machine-readable discovery files (llms.txt) confirmed; onboarding relies on manual CLI setup.",
    "Sandboxed review/approval workflow, while safe, adds friction to first-try success for autonomous agents."
  ],
  "recommendations": [
    "Publish a stable, resolvable canonical domain and verify DNS before relying on it as a recommended resource.",
    "Add an llms.txt and clear robots.txt so agents can safely crawl and summarize the platform.",
    "Expose a documented, structured API or integration endpoints to complement the CLI for agent-driven workflows.",
    "Provide a magic-link or identity-provider-based sign-up to reduce onboarding friction.",
    "Maintain a public status page and uptime signal so consistency can be assessed objectively."
  ]
}