{
  "fixes": {
    "slug": "lm-studio",
    "name": "LM Studio",
    "listing": "https://www.anchorterminal.com/tools/lm-studio",
    "markdown": "# Fix list: LM Studio\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/lm-studio, the October 2026 research run, assessed 3 October 2026. Grade C, 57.9 out of 100.\n\nThis is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.\n\nFor a coding agent working on LM Studio: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.\n\n## 1. Reliability, 34 out of 100, up to 13.2 more on the total\n\nWhy it scored 34: Read with the local-software lines, since LM Studio and llmster run on the owner's machine with no hosted service behind the API. Installers for macOS 14 or later on Apple silicon, Windows x64 (AVX2) and ARM, and a Linux AppImage for x64 and ARM64 on Ubuntu 20.04 or later, with requirements on one page, plus a one-line install script for llmster. No package-manager route, and the docs say the Linux in-app updater is still in the works (18). The app and llmster are closed source, so there's no public CI or test suite, and the public `lms` and lmstudio-js repositories run only a CLA workflow (0). The public bug tracker shows 1.7k open issues. A July crash report (#2149, a stack buffer overrun in llama-server.exe) got a staff reply asking for dumps and is still open, and the newest open issues our reader saw dated from mid-July (8). Versions run 0.4.x with dated notes per release, and the API changelog flags behaviour changes (0.3.23 moved gpt-oss reasoning out of `message.content`), but it stops at 0.4.1 with no dates after 0.3.29, while 0.4.22 and 0.4.24 changed API behaviour (8). 0.4.25, pre-1.0. The docs call the v1 REST API officially released, which we don't count as a stable declaration for the product (0).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):\n\nHosted APIs, MCP servers, models and platforms.\n\n- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.\n\nLocal packages, SDKs, frameworks and stdio MCP servers.\n\n- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.\n\nProtocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.\n\n## 2. Security \u0026 auth, 59 out of 100, up to 7.2 more on the total\n\nWhy it scored 59: Read with the tool checklist. Named API tokens (`sk-lm-` prefix) with permissions picked at creation, shown once, editable and deletable, sent in a header. Authentication is off by default, so any local process can call the server until the owner turns it on, and the server binds to localhost unless Serve on Local Network is on (24). Token permissions, `allowed_tools` per integration, and two separate switches before the API can reach MCP servers (remote servers named per request, and the owner's mcp.json servers, which also need authentication on). The app asks before each MCP tool call with editable arguments, but tools called through the API run without a prompt (15). The docs warn that MCP servers can run code and read files and say to install none from untrusted sources. Nothing on injected instructions in tool results or documents (6). `lms log stream --source server` and `--source model` show each request and the model's input and output, with `--json` (12). No security.txt (lmstudio.ai answers that path with a Hub web page), no SECURITY.md in the public repositories, and no disclosure policy, bounty or certification found. The changelog's \"Security hardening\" lines in 0.4.16 and 0.4.17 say nothing more, and NVD lists no CVE for LM Studio (2).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-security):\n\n- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.\n\nModels are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.\n\n## 3. Schema \u0026 documentation, 64 out of 100, up to 5.9 more on the total\n\nWhy it scored 64: Read with the API lines. No OpenAPI file or other machine-readable contract for the server that we found. Each REST page carries a typed parameter table, the TypeScript SDK defines its types in zod, and the compatible endpoints follow OpenAI's and Anthropic's shapes (8). An llms.txt of about 2,850 lines at lmstudio.ai that, in what our reader saw, describes the 0.3 app and doesn't mention the v1 REST API, API tokens, llmster or MCP. The docs source is public Markdown on GitHub (6). The REST overview compares /api/v1/chat with /v1/responses, /v1/chat/completions and /v1/messages feature by feature, and the server settings say which switches carry risk (\"This can be a security risk if you've defined MCP servers that have access to your file system or private data\") (15). Parameter tables give types, optional flags and enumerated values (download status is one of five states), and the error object has a `type` with eight values such as invalid_request, model_not_found and mcp_connection_error, plus `message`, `code` and `param` (12). curl, Python and TypeScript examples on the REST pages, and documented error and streaming event shapes (12). Versioned paths (/api/v0, then /api/v1 since 0.4.0) and a dated changelog per release at /changelog/lmstudio, with the API changelog lagging as noted (11).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):\n\nAPIs and MCP servers.\n\n- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.\n\nModels are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.\n\n## 4. Agent ergonomics, 69 out of 100, up to 5 more on the total\n\nWhy it scored 69: Read with the API lines. `/api/v1/chat` keeps state on the server, so a caller sends `previous_response_id` instead of the whole history, `allowed_tools` trims the MCP tool list a model sees, and `max_output_tokens` and `context_length` are set per request (20). Model listing has no paging. `lms ls --json` and `lms ps --json` give machine-readable lists, and `lms log stream` filters by source and prints JSON (12). Typed error objects with a type, message, code and parameter, and a stream emits an `error` event and still ends with `chat.end` (16). Chat completions are stateless and safe to repeat, downloads return a job id to poll, and `store: false` turns off stored chats per request. No retry or idempotency guidance (8). A request naming a downloaded model loads it just in time, with idle-unload settings, and official SDKs exist for TypeScript (2.0.0) and Python, though Python's last stable release is from August 2025 (13).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):\n\n- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.\n\nModels are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.\n\n## 5. Payments \u0026 pricing, 60 out of 100, up to 5 more on the total\n\nWhy it scored 60: Self-hosted rule, scored for what an agent uses, the local server in the LM Studio app or llmster. No x402, MPP or L402 (0). The app, llmster and the server are free for personal and work use with no account or card, so 20, 20 and 20 on the last three lines. Element Labs sells cloud model plans ($20 and $100 a month) for Bionic, a separate app this listing doesn't cover.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):\n\nThe published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).\n\n- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.\n- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for \"contact sales\" or prices behind a login.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).\n\nPayment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.\n\nOpen-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.\n\n## 6. Transparency \u0026 trust, 61 out of 100, up to 3.4 more on the total\n\nMade of editorial 55, provenance 67.\n\nWhy it scored 61: The app and llmster are proprietary, under terms effective 23 August 2026 that allow personal and internal business use and forbid reverse engineering and redistribution. `lms` and the TypeScript and Python SDKs are MIT (17). The privacy policy (effective June 2026) names Element Labs, Inc., says messages, chat histories and documents never leave the machine with local models, lists update checks (app version, OS, IP address) and anonymised model-search queries, and says cloud requests aren't kept. The offline page agrees. Retention is \"as long as necessary\" with no periods, processors are named only by category, and LM Link's use of Tailscale is in the docs but not the policy (18). The v0 REST API stays documented beside v1, but there's no deprecation policy, and the terms let Element Labs change, suspend or discontinue parts of the software with no stated notice (6). No analytics described, and what the app sends is listed in the policy. The only way to stop the update check is to stay offline, and with the source closed we couldn't confirm the list (14).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):\n\n- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).\n\nThe other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.\n\nProvenance checks not met in full (half of this category, computed from checked facts):\n\n- Domain age: lmstudio.ai, registered 2023-05-03 (3 years) (7 of 15)\n- Status page: not found (0 of 10)\n- security.txt: not found (0 of 10)\n\n## 7. Maintenance \u0026 community, 72 out of 100, up to 2.5 more on the total\n\nWhy it scored 72: 0.4.25 on 19 September 2026 (30). Seven releases since 5 July, 0.4.19 to 0.4.25 (20). Closed-source app, so the 0 to 15 scale for closed services. A dated changelog per release, a public bug tracker where staff answered the July crash report we read, a Discord and bugs@lmstudio.ai, against 1.7k open issues (10). Official SDKs for TypeScript (@lmstudio/sdk 2.0.0, prepared on 21 September 2026) and Python, whose last stable release (1.5.0, 22 August 2025) predates API tokens (9). The public `lms` and lmstudio-js repositories run no build or test workflow, and the Python SDK's repository, which has one, hasn't had a commit since 26 October 2025 (3).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):\n\n- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.\n\nModels are read for deprecation notice periods and model churn rather than release counts.\n\n## What we couldn't check\n\nWhat we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.\n\n- Whether the bug tracker has newer open issues than the mid-July ones our reader saw, and how fast staff reply\n- Which permissions an API token can carry. The docs show them only in screenshots\n- Whether Allow per-request MCPs is off by default. The docs say it must be enabled but don't give the default\n- How a headless llmster owner creates API tokens. The docs describe only the app's Developer page\n- Unchecked: what the app and llmster send besides the update check and model search, since the source isn't public\n- Whether the llms.txt we read was complete. Our reader saw about 2,850 lines describing the 0.3 app\n\n## Weaknesses\n\n- Authentication is off by default, so any local process can call the server\n- Closed-source app and daemon with no public CI or test suite\n- The Python SDK's last stable release (1.5.0, 22 August 2025) can't send API tokens, and the docs name an environment variable no release reads\n- No OpenAPI file, and the llms.txt we read covers the 0.3 app, not the v1 REST API, tokens, llmster or MCP\n- 1.7k open issues in the public bug tracker\n\n## What costs an agent a turn today\n\nThe notes we give agents before they call it. Each one is a workaround an agent shouldn't need.\n\n- Send `Authorization: Bearer $LM_API_TOKEN` when the owner gives you a token. With Require Authentication on, every request needs it\n- Pass `previous_response_id` to /api/v1/chat instead of resending the history, and `store: false` for one-off calls\n- List models with `GET /api/v1/models` before naming one. A named model that's downloaded loads just in time\n- Install the Python SDK pre-release (1.6.0b1) and pass `api_token` directly. It reads `LMSTUDIO_API_TOKEN`, not the `LM_API_TOKEN` the docs name\n- Set `allowed_tools` on every MCP integration. Without it the model sees every tool on the server\n\n## What the review panel asked for\n\n- a current API changelog\n- a stable Python SDK release\n- publish a security policy\n- document token permissions\n\n## When it's done\n\nSend what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `\"kind\": \"dispute\"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.\n",
    "grade": "C",
    "score": 57.9,
    "assessed": "2026-10-03",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 34,
        "maxGain": 13.2,
        "reason": "Read with the local-software lines, since LM Studio and llmster run on the owner's machine with no hosted service behind the API. Installers for macOS 14 or later on Apple silicon, Windows x64 (AVX2) and ARM, and a Linux AppImage for x64 and ARM64 on Ubuntu 20.04 or later, with requirements on one page, plus a one-line install script for llmster. No package-manager route, and the docs say the Linux in-app updater is still in the works (18). The app and llmster are closed source, so there's no public CI or test suite, and the public `lms` and lmstudio-js repositories run only a CLA workflow (0). The public bug tracker shows 1.7k open issues. A July crash report (#2149, a stack buffer overrun in llama-server.exe) got a staff reply asking for dumps and is still open, and the newest open issues our reader saw dated from mid-July (8). Versions run 0.4.x with dated notes per release, and the API changelog flags behaviour changes (0.3.23 moved gpt-oss reasoning out of `message.content`), but it stops at 0.4.1 with no dates after 0.3.29, while 0.4.22 and 0.4.24 changed API behaviour (8). 0.4.25, pre-1.0. The docs call the v1 REST API officially released, which we don't count as a stable declaration for the product (0).",
        "checklist": [
          "Hosted APIs, MCP servers, models and platforms.",
          "- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.",
          "Local packages, SDKs, frameworks and stdio MCP servers.",
          "- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.",
          "Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-reliability"
      },
      {
        "key": "security",
        "name": "Security \u0026 auth",
        "score": 59,
        "maxGain": 7.2,
        "reason": "Read with the tool checklist. Named API tokens (`sk-lm-` prefix) with permissions picked at creation, shown once, editable and deletable, sent in a header. Authentication is off by default, so any local process can call the server until the owner turns it on, and the server binds to localhost unless Serve on Local Network is on (24). Token permissions, `allowed_tools` per integration, and two separate switches before the API can reach MCP servers (remote servers named per request, and the owner's mcp.json servers, which also need authentication on). The app asks before each MCP tool call with editable arguments, but tools called through the API run without a prompt (15). The docs warn that MCP servers can run code and read files and say to install none from untrusted sources. Nothing on injected instructions in tool results or documents (6). `lms log stream --source server` and `--source model` show each request and the model's input and output, with `--json` (12). No security.txt (lmstudio.ai answers that path with a Hub web page), no SECURITY.md in the public repositories, and no disclosure policy, bounty or certification found. The changelog's \"Security hardening\" lines in 0.4.16 and 0.4.17 say nothing more, and NVD lists no CVE for LM Studio (2).",
        "checklist": [
          "- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.",
          "Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-security"
      },
      {
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "score": 64,
        "maxGain": 5.9,
        "reason": "Read with the API lines. No OpenAPI file or other machine-readable contract for the server that we found. Each REST page carries a typed parameter table, the TypeScript SDK defines its types in zod, and the compatible endpoints follow OpenAI's and Anthropic's shapes (8). An llms.txt of about 2,850 lines at lmstudio.ai that, in what our reader saw, describes the 0.3 app and doesn't mention the v1 REST API, API tokens, llmster or MCP. The docs source is public Markdown on GitHub (6). The REST overview compares /api/v1/chat with /v1/responses, /v1/chat/completions and /v1/messages feature by feature, and the server settings say which switches carry risk (\"This can be a security risk if you've defined MCP servers that have access to your file system or private data\") (15). Parameter tables give types, optional flags and enumerated values (download status is one of five states), and the error object has a `type` with eight values such as invalid_request, model_not_found and mcp_connection_error, plus `message`, `code` and `param` (12). curl, Python and TypeScript examples on the REST pages, and documented error and streaming event shapes (12). Versioned paths (/api/v0, then /api/v1 since 0.4.0) and a dated changelog per release at /changelog/lmstudio, with the API changelog lagging as noted (11).",
        "checklist": [
          "APIs and MCP servers.",
          "- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.",
          "Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-schema"
      },
      {
        "key": "ergonomics",
        "name": "Agent ergonomics",
        "score": 69,
        "maxGain": 5,
        "reason": "Read with the API lines. `/api/v1/chat` keeps state on the server, so a caller sends `previous_response_id` instead of the whole history, `allowed_tools` trims the MCP tool list a model sees, and `max_output_tokens` and `context_length` are set per request (20). Model listing has no paging. `lms ls --json` and `lms ps --json` give machine-readable lists, and `lms log stream` filters by source and prints JSON (12). Typed error objects with a type, message, code and parameter, and a stream emits an `error` event and still ends with `chat.end` (16). Chat completions are stateless and safe to repeat, downloads return a job id to poll, and `store: false` turns off stored chats per request. No retry or idempotency guidance (8). A request naming a downloaded model loads it just in time, with idle-unload settings, and official SDKs exist for TypeScript (2.0.0) and Python, though Python's last stable release is from August 2025 (13).",
        "checklist": [
          "- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.",
          "Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-ergonomics"
      },
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 60,
        "maxGain": 5,
        "reason": "Self-hosted rule, scored for what an agent uses, the local server in the LM Studio app or llmster. No x402, MPP or L402 (0). The app, llmster and the server are free for personal and work use with no account or card, so 20, 20 and 20 on the last three lines. Element Labs sells cloud model plans ($20 and $100 a month) for Bionic, a separate app this listing doesn't cover.",
        "checklist": [
          "The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).",
          "- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.\n- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for \"contact sales\" or prices behind a login.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).",
          "Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.",
          "Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-payments"
      },
      {
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 61,
        "maxGain": 3.4,
        "reason": "The app and llmster are proprietary, under terms effective 23 August 2026 that allow personal and internal business use and forbid reverse engineering and redistribution. `lms` and the TypeScript and Python SDKs are MIT (17). The privacy policy (effective June 2026) names Element Labs, Inc., says messages, chat histories and documents never leave the machine with local models, lists update checks (app version, OS, IP address) and anonymised model-search queries, and says cloud requests aren't kept. The offline page agrees. Retention is \"as long as necessary\" with no periods, processors are named only by category, and LM Link's use of Tailscale is in the docs but not the policy (18). The v0 REST API stays documented beside v1, but there's no deprecation policy, and the terms let Element Labs change, suspend or discontinue parts of the software with no stated notice (6). No analytics described, and what the app sends is listed in the policy. The only way to stop the update check is to stay offline, and with the source closed we couldn't confirm the list (14).",
        "blend": "editorial 55, provenance 67",
        "checklist": [
          "- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).",
          "The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-transparency"
      },
      {
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 72,
        "maxGain": 2.5,
        "reason": "0.4.25 on 19 September 2026 (30). Seven releases since 5 July, 0.4.19 to 0.4.25 (20). Closed-source app, so the 0 to 15 scale for closed services. A dated changelog per release, a public bug tracker where staff answered the July crash report we read, a Discord and bugs@lmstudio.ai, against 1.7k open issues (10). Official SDKs for TypeScript (@lmstudio/sdk 2.0.0, prepared on 21 September 2026) and Python, whose last stable release (1.5.0, 22 August 2025) predates API tokens (9). The public `lms` and lmstudio-js repositories run no build or test workflow, and the Python SDK's repository, which has one, hasn't had a commit since 26 October 2025 (3).",
        "checklist": [
          "- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.",
          "Models are read for deprecation notice periods and model churn rather than release counts."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-maintenance"
      }
    ],
    "provenance": [
      {
        "label": "Domain age",
        "value": "lmstudio.ai, registered 2023-05-03 (3 years)",
        "points": 7,
        "max": 15
      },
      {
        "label": "Status page",
        "value": "not found",
        "points": 0,
        "max": 10
      },
      {
        "label": "security.txt",
        "value": "not found",
        "points": 0,
        "max": 10
      }
    ],
    "unchecked": [
      "Whether the bug tracker has newer open issues than the mid-July ones our reader saw, and how fast staff reply",
      "Which permissions an API token can carry. The docs show them only in screenshots",
      "Whether Allow per-request MCPs is off by default. The docs say it must be enabled but don't give the default",
      "How a headless llmster owner creates API tokens. The docs describe only the app's Developer page",
      "Unchecked: what the app and llmster send besides the update check and model search, since the source isn't public",
      "Whether the llms.txt we read was complete. Our reader saw about 2,850 lines describing the 0.3 app"
    ],
    "weaknesses": [
      "Authentication is off by default, so any local process can call the server",
      "Closed-source app and daemon with no public CI or test suite",
      "The Python SDK's last stable release (1.5.0, 22 August 2025) can't send API tokens, and the docs name an environment variable no release reads",
      "No OpenAPI file, and the llms.txt we read covers the 0.3 app, not the v1 REST API, tokens, llmster or MCP",
      "1.7k open issues in the public bug tracker"
    ],
    "agentNotes": [
      "Send `Authorization: Bearer $LM_API_TOKEN` when the owner gives you a token. With Require Authentication on, every request needs it",
      "Pass `previous_response_id` to /api/v1/chat instead of resending the history, and `store: false` for one-off calls",
      "List models with `GET /api/v1/models` before naming one. A named model that's downloaded loads just in time",
      "Install the Python SDK pre-release (1.6.0b1) and pass `api_token` directly. It reads `LMSTUDIO_API_TOKEN`, not the `LM_API_TOKEN` the docs name",
      "Set `allowed_tools` on every MCP integration. Without it the model sees every tool on the server"
    ],
    "requests": [
      {
        "text": "a current API changelog",
        "reviews": 1
      },
      {
        "text": "a stable Python SDK release",
        "reviews": 1
      },
      {
        "text": "publish a security policy",
        "reviews": 1
      },
      {
        "text": "document token permissions",
        "reviews": 1
      }
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-04",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.3",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run"
  }
}