MCP-сервер app.nz

• Адресhttps://app.nz/mcpнет в официальном реестре: ответил на запрос

Ни одна запись официального реестра MCP не называет этот сервер. Он ответил, когда Атлас запросил app.nz по одному из обычных для MCP-серверов адресов, значит, сервер принадлежит самому вендору. Клиенты и каталоги, которые берут серверы из реестра, его не показывают.

• Состояниеотвечаетпроверено 2026-10-28
• Инструменты93
• Сервер называет себяapp.nz
• Версия сервера1.0.0
• Версия протокола2025-06-18

2026-10-28 сервер ответил на initialize и tools/list без входа. Первый ответ за 93 мс.

• Инструменты делают все проверенные сценарии: hand a coding task to an agent that opens a pull request; deploy a GPU model endpoint that scales to zero.
• Проверок пока 6 из 7, нужных, чтобы судить о надёжности.

Вывод сделан правилами по замерам выше. Вероятный первый вызов выбрала модель по списку инструментов: прочитайте описание, прежде чем на него полагаться.

• Оценка INITE MCPчерез неделюдля оценки нужно 7 ежедневных проверок, пока 6
• Что агент может с ним сделать2 из 2 сценариев
• Отвечалидёт замерпроверок пока 6, долю покажем после 7
• Медиана первого ответа166 мс
• Место в контексте агента≈ 10 070 токеновописания всех инструментов (93), они занимают контекст агента на каждом ходу
• Описания инструментовC · 74 из 100
• Типизированные ответы0 из 93 инструментовобъявлен outputSchema: ответ приходит с типами и читается сразу

• Параметры без описания: в 54 из 93
• Не сказано, только читает инструмент или может удалять: 93 из 93

Программа проверяет сервер раз в день: initialize и tools/list, без модели. Токены - это длина описаний из расчёта примерно четыре символа на токен. Оценка - среднее шести проверок имени, описания, схемы и аннотаций каждого инструмента. Она говорит, насколько понятно описаны инструменты, а не о том, работают ли они. Оценка складывается только из замеренного: отвечает ли сервер (25%), делает ли работу продукта (25%), описания (20%), контекст (10%), использование (10%) и насколько просто подключиться (10%). Незамеренная часть не учитывается, а скрытая инструкция ограничивает оценку 40 баллами. Состояние MCP-серверов вендоров: методика и цифры →

• chat

  Run a chat completion against any app.nz model (OpenAI, Anthropic, Google, xAI, and more). Returns the assistant's text reply.

• run_fusion

  Run one prompt through a panel of models in parallel, then a judge model regenerates the strongest final answer from every candidate. Accepts any gateway model id including custom OpenRouter slugs (openrouter/vendor/model).

• list_models

  List routable app.nz models with provider, modality and context window. Optional case-insensitive substring filter.

• monitoring_status

  Read the latest app.nz Git, Cog, and site-hosting SRE probes, six-hour remediation policy, and your site-monitor rollup.

• list_marketplace_apps

  Browse games, tools, education products, and community projects in the public app.nz apps marketplace.

• publish_marketplace_app

  Publish a hosted or external app immediately to app.nz/apps. There is no approval queue; title, description, and OG image are discovered from the URL when omitted.

• generate_image

  Generate an image from a text prompt and return image URL(s). Use model 'seamless-tiling' for edge-compatible game terrain and repeating patterns.

• run_texture_agent

  Generate a bounded set of eBank or other seamless-texture candidates, add an exact procedural control, verify every 2x2 repeat, repair conservative seam failures, and return a pollable experiment.

• get_texture_agent

  Poll a texture experiment and return every candidate, verification metrics, 2x2 proof, and the highest-scoring accepted tile.

• generate_panorama

  Convert a reference image into an equirectangular 360 panorama or a spatial world file with Hunyuan World.

• list_comfy_tutorials

  List the 10-part ComfyUI course. Every lesson includes a real gallery workflow slug, blog path, concepts, duration, and downloadable API JSON.

• list_startup_growth_series

  List the fully voiced 10-part Growth Engineering series for startups and traditional businesses, including reading and prerecorded audio URLs.

• get_startup_growth_chapter

  Read one long-form Growth Engineering chapter with its stable blog and OmniServe-generated audio URLs.

• list_comfy_labs

  List the 12 Creative Systems Labs for audio, talking characters, reactive animation, style, frame transforms, structural control, and 3D. Every lab points to runnable JSON and open-source integrations.

• get_comfy_workflow

  Get a vetted gallery workflow with API-format JSON, named inputs, model sources, VRAM, provenance, verification, and tutorial links.

• deploy_comfy_gallery_workflow

  Deploy a vetted ComfyUI gallery workflow with its tested named inputs and minimum VRAM contract.

• deploy_comfy_workflow

  Deploy a ComfyUI API-format workflow to an app.nz remote GPU. Use gpu-rtx4090 with mode pod for an exact 24GB RTX 4090; serverless scales to zero.

• run_comfy_workflow

  Queue inference on a deployed ComfyUI remote GPU workflow and return a run id for polling.

• get_comfy_run

  Get remote ComfyUI inference status, runtime, cost, tier, errors, and output URLs.

• list_music_pipelines

  List open music datasets, reproducible Open Piano model previews, license posture, piano/RL/diffusion trainers, bounded Opus/R2 storage, and Cog/RunPod deployment targets.

• generate_audio

  Generate music or sound effects from a prompt, save it to the indexed audio library, and return the audio URL.

• search_audio

  Search the generated audio library with semantic search when enabled and SQL keyword fallback.

• embed

  Create embedding vectors for text input.

• search_mcp_servers

  Search the app.nz MCP directory — a curated index of Model Context Protocol servers (connectors for Gmail, GitHub, databases, payments, and more). Returns matching servers with their install/transport details.

• mail_search

  Search the caller's app.nz mailbox using exact operators and optional Gobed hybrid semantic ranking.

• mail_read

  Read one owned message, including unsubscribe availability and receiving-edge authenticity metadata.

• mail_received_read

  Read-only inspection of one received message, including unsubscribe availability and receiving-edge authenticity metadata. Rejects sent messages and performs no external action.

• mail_unsubscribe_intent

  Prepare a user-confirmable unsubscribe intent. Never visits sender URLs, submits one-click requests, or sends mail; the returned target is sender-provided and performed is always false.

• mail_send

  Send a human-reviewed message from an owned app.nz mailbox.

• mail_rules_list

  List automatic mailbox organization and forwarding rules.

• mail_rule_create

  Create a rule that labels, categorizes, marks important/read, or forwards matching mail.

• mail_contacts_list

  List or search the caller's permission-aware email contacts and subscription states.

• mail_contact_create

  Create or update one permission-aware email contact. Existing unsubscribed or suppressed contacts are never silently resubscribed.

• mail_audiences_list

  List the caller's campaign audiences and current member counts.

• mail_audience_create

  Create an empty campaign audience. This does not send email.

• mail_templates_list

  List saved email templates, including compiled HTML and plain text.

• mail_campaigns_list

  List draft campaigns and current permission-aware recipient counts. Listing never sends email.

• mail_design_generate

  Use the mail design agent to produce validated campaign copy plus server-compiled, email-safe HTML and plain text.

• search_web

  Search the live web, the compact WebFiddle index, or run the generic app.nz search agent. Indexed results include readable cached-page links.

• deep_research

  Start a multi-source cited research run using live web, compact index, papers, and cached Markdown reading.

• search_papers

  Semantically search the app.nz research index for papers, methods, or datasets, including paper metadata and linked code. Uses one paper-search credit.

• read_page

  Fetch a public webpage as cached, readable Markdown. Requests Markdown from compatible origins/Cloudflare first and otherwise performs Lynx-style HTML extraction.

• create_doc_artifact

  Create a first-class document artifact. Data is Markdown text and opens in /write.

• update_doc_artifact

  Update one of your document artifacts. Data is Markdown text and opens in /write.

• create_sheet_artifact

  Create a first-class spreadsheet artifact. Data is JSON grid data and opens in /sheets.

• update_sheet_artifact

  Update one of your spreadsheet artifacts. Data is JSON grid data and opens in /sheets.

• list_datasets

  List your uploaded table, database, image, and video datasets. Returns ids, formats, dimensions/row counts, and streamable file URLs.

• get_dataset

  Get one of your datasets by id or slug, including schema/sample rows for tables or a streamable URL for image/video data.

• search_creation_feed

  Search app.nz creator posts for tutorials, code, models, datasets, artwork, videos, research, creator handles, and attached artifacts. Uses hybrid Gobed semantic ranking when available.

• publish_creation_post

  Publish a model, dataset, coding artifact, narrated tutorial, research note, artwork, video, or general post to the authenticated creator's app.nz feed.

• like_creation_post

  Like or unlike a public app.nz creation-feed post.

• start_manim_course_agent

  Launch a production narrated Manim course on a Hetzner CPU worker using GPT-5.6 Sol with medium reasoning. Defaults to 20 lessons and publishes videos, speech, captions, posters, source, tests, visualbench evidence, and creator posts.

• run_agent

  Launch an app.nz coding agent with Auto, Codex, Grok, Claude Code, Gemini, or Cursor. Set mode to goal, plan, review, fix, test, explain, or deploy, or begin the prompt with the matching slash command.

• list_goals

  List the authenticated user's goals with progress, milestones, and check-in history.

• create_goal

  Create a measurable goal with a target, current value, due date, priority, and optional milestones.

• update_goal

  Update goal fields or status. Status is active, paused, completed, or archived.

• check_in_goal

  Record a new metric value and note on a goal, preserving its progress history.

• ci_list_pipelines

  List your app.nz CI pipelines (portable app.yaml job DAGs plus GitHub Actions, GitLab, CircleCI, Buildkite, and Bamboo configs) with run and success counts.

• ci_migrate_config

  Convert a GitHub Actions, GitLab, CircleCI, Buildkite, or Bamboo CI config into portable app.yaml. Optionally launch a Codex migration agent that inspects the repo, validates the result, and opens a reviewable PR.

• ci_run_pipeline

  Trigger a run of one of your CI pipelines. Jobs execute as a DAG; poll ci_get_run for status and search logs with ci_search_logs.

Первые 60 из 93, в порядке сервера.