{
  "schema_version": "2.0",
  "slug": "gptme-gptme",
  "name": "gptme/gptme",
  "agent_url": "https://github.com/gptme/gptme",
  "category": "Coding",
  "run_id": "run-gptme-gptme-u0tgstbd",
  "run_at": "2026-06-13T06:00:59.870Z",
  "editor": "Hlido Editor",
  "editorial_method": "public-surface-tier-1+editorial-narrative-v2",
  "methodology_version": "2026.07",
  "methodology_url": "/methodology/public-surface-tier-1/",
  "score": 75,
  "tier": "STEADY",
  "laddoo_score": 75,
  "confidence": "medium",
  "hlido_opinion": {
    "headline": "One of the original terminal coding agents — mature, provider-agnostic, and genuinely scriptable, if you're comfortable living in the CLI.",
    "body": "gptme has been around since spring 2023, which in agent-tool years makes it an elder, and the maturity shows: 111 releases, 4,000+ commits, and a design that treats the terminal as the whole product rather than a fallback. It writes code, runs shell and Python, browses via Playwright, and reads images — all locally, with your own keys, against Anthropic, OpenAI, Google, DeepSeek, OpenRouter or a local llama.cpp. The part that matters for an agentic ecosystem is that it was built to be driven, not just chatted with: `--non-interactive`, `--no-confirm`, and `--output-format json` make it a legitimate CI/automation primitive, and it speaks both MCP (as a client) and ACP (as a server for editor embedding). It positions itself against Claude Code, Cursor and Warp, and on the axis of 'open, local-first, self-hostable' it holds that ground honestly. Where it asks something of you: this is a developer tool with a developer's onboarding — no hosted UI to hide behind, the polish is functional rather than designed, and getting value out of it assumes you already think in terminals and pipx. That's a deliberate trade, not a defect.",
    "voice": "Hlido Editor",
    "as_of": "2026-07-21",
    "editor_signature_pending": true
  },
  "tier_rationale": "STEADY (75) because the tool is demonstrably mature and actively maintained (v0.32.1, 111 releases), the MIT license and local-first architecture are clean trust signals, and the automation surface (JSON output, non-interactive mode, MCP/ACP) is real rather than aspirational. Not higher because the public surface is a GitHub README rather than a product with an evidenced onboarding path, and the experience is unapologetically CLI-only — excellent for its audience, narrow for everyone else.",
  "what_it_does_well": [
    "Runs as a genuine CLI primitive — non-interactive mode and JSON output make it CI/automation-friendly",
    "Provider-agnostic: Anthropic, OpenAI, Google, xAI, DeepSeek, OpenRouter, or local llama.cpp",
    "Local-first and self-hostable — your keys, your machine, no mandatory cloud",
    "Speaks MCP (client) and ACP (server) so it composes with other tools and editors",
    "Long, consistent maintenance track record (since 2023, 100+ releases)"
  ],
  "what_it_fails_at": [
    "CLI-only onboarding — no hosted UI, assumes terminal + pipx fluency",
    "Public surface is a README; there is no product marketing making verifiable claims to audit",
    "Polish is functional, not designed — rough edges are part of the deal",
    "Web UI and desktop app are secondary/in-development rather than first-class"
  ],
  "best_for": [
    "Developers who want an open, scriptable coding agent they can run in CI",
    "Teams that need provider flexibility and local-first execution for privacy or cost",
    "Anyone embedding an agent into an editor via ACP or a pipeline via MCP",
    "Users who prefer an auditable open-source tool over a hosted black box"
  ],
  "not_recommended_for": [
    "Non-technical users who need a polished GUI",
    "Buyers who want vendor support and SLAs (this is community open source)",
    "Anyone unwilling to manage their own API keys and environment"
  ],
  "red_flags": [],
  "compared_to": [
    {
      "slug": "aider",
      "verdict_diff": "Aider is the tighter git-native pair-programmer focused on edit-commit loops; gptme is the broader terminal agent that also runs shell/Python and browses. Pick Aider for disciplined in-repo edits, gptme for general terminal knowledge-work.",
      "preferred_for_axis": "breadth-of-terminal-tooling"
    },
    {
      "slug": "goose",
      "verdict_diff": "Goose and gptme occupy similar 'agent in your terminal' territory; gptme's edge is its longer maturity and explicit JSON/non-interactive automation surface, Goose's is its extension ecosystem.",
      "preferred_for_axis": "automation-scriptability"
    }
  ],
  "evidence_urls": [
    {
      "claim": "Terminal agent that writes code, uses the shell, browses the web",
      "source": "https://github.com/gptme/gptme",
      "tested_at": "2026-07-21",
      "verified": true
    },
    {
      "claim": "MIT licensed, 4.4k stars, 111 releases (v0.32.1)",
      "source": "https://github.com/gptme/gptme",
      "tested_at": "2026-07-21",
      "verified": true
    },
    {
      "claim": "Provider-agnostic incl. local llama.cpp; MCP + ACP support",
      "source": "https://github.com/gptme/gptme",
      "tested_at": "2026-07-21",
      "verified": true
    },
    {
      "claim": "Non-interactive mode and --output-format json for CI",
      "source": "https://github.com/gptme/gptme",
      "tested_at": "2026-07-21",
      "verified": true
    }
  ],
  "agent_relevance": {
    "has_api": true,
    "has_cli": true,
    "has_mcp": true,
    "has_webhook": false,
    "has_sdk": false,
    "behavioral_testable": true,
    "agent_integration_path": "Strong. gptme is designed to be driven programmatically: a built-in REST server, JSON-line output, --non-interactive/--no-confirm flags, MCP client support and an ACP server for editor embedding. It can be a component inside a larger agent pipeline, not just a human chat tool.",
    "agent_friendly_score": 8
  },
  "marking_signal": {
    "marking_statement": false,
    "detection_tool": false,
    "cop_signatory": null,
    "evidence_url": null,
    "checked_at": "2026-07-25",
    "source": "engine-browser-marking-retrofit"
  },
  "checklist": [
    {
      "id": "homepage_loads",
      "pass": true,
      "required": true
    },
    {
      "id": "primary_value_prop",
      "pass": true,
      "required": true
    },
    {
      "id": "cta_present",
      "pass": false,
      "required": true
    },
    {
      "id": "pricing_or_access",
      "pass": false,
      "required": false
    },
    {
      "id": "evidence_or_demo",
      "pass": true,
      "required": false
    }
  ],
  "summary": "One of the original terminal coding agents — mature, provider-agnostic, and genuinely scriptable, if you're comfortable living in the CLI.",
  "_summary_deprecation_note": "Field kept as a v1-compatibility alias of hlido_opinion.headline. New consumers should read hlido_opinion.{headline,body,voice,as_of}.",
  "staleness_after": "2026-10-19",
  "review_age_days_at_publish": 38,
  "next_review_due_at": "2026-10-19",
  "attestation_url": "/data/attestations/gptme-gptme.json",
  "signature_pending": true,
  "source": "hlido-editor-v2",
  "evidence_images": {
    "run_id": "run-cf6955366bff1626-github-com",
    "base": "https://images.hlido.eu/reviews/gptme-gptme/run-cf6955366bff1626-github-com",
    "files": [
      "home.png"
    ],
    "urls": [
      "https://images.hlido.eu/reviews/gptme-gptme/run-cf6955366bff1626-github-com/home.png"
    ],
    "verified_at": "2026-07-31T09:31:38.575Z",
    "verified_count": 1,
    "note": "Screenshots captured by the Hlido engine during the reviewed run, served from R2. Every URL here was HEAD-verified at the timestamp above. `run_id` is the ENGINE run id — it differs from `scorecard.run_id` and is the only one these keys resolve under."
  }
}