All servers / Other / Zion MCP server
Zion MCP server
MCP server for functorz.com
Tested by ConnectATDScore 77 out of 100. Tested 2026-11-19 on package version 2.8.5. Package: zion-mcp
Score breakdown
| Runs | 18 / 20 | |
|---|---|---|
| Tool descriptions | 15 / 20 | |
| Package safety | 16 / 20 | |
| Maintenance | 20 / 20 | |
| Adoption | 8 / 20 |
What we saw
- Starts and answers the MCP handshake: yes (2.1 seconds including download)
- Tools offered: 49
- Last published: 2026-11-18
- Weekly downloads (npm): 598
- License: MIT · Direct dependencies: 2
Want this server set up for your team?
ConnectATD builds AI automation and wires up MCP servers like this one. Tell us what you are trying to do.
Talk to ConnectATDTools and how they are described
Agents choose tools from these descriptions. Under each name you can see how many of its parameters are explained.
list_actionflow_node_templatesno parametersList the preset action-flow integration nodes (the TEMPLATE_CODE catalog — e.g. SMS, video/AI generation) for planning: each template’s templateCodeId, displayName, group, async flag, and input/output param types. To add one, read schema_actionflow_node op=BROWSE_ACTION_FLOW_ADD_NODE_OPTIONS at the position and pass that entry’s `node` to op=ADD_NODE. A TEMPLATE_CODE node already in the schema carries one of these templateCodeIds; never rewrite it to an id from anywhere else.
get_tab_bar_iconsno parametersThe preset tab-bar icons this deployment ships, as name and media exId — home, cart, chat, profile and the rest. A tab that is shown needs an icon for both its normal and its selected state or the project reports an error, and the exIds are hashids of media rows that cannot be derived from an icon’s name, so take them from here before schema_tab_bar op=UPDATE_TAB_BAR_ITEMS — that tool rejects a value that could not be an exId rather than letting the mini-program build fail over a file it cannot find. To use an icon that is not in the list, upload it in the editor and use the exId it is stored under.
set_current_working_directory0 of 1 parameter describedSet the workspace root for local credentials, daemon registry, and project context.
get_account_profileno parametersFetch current account and organization/project-space metadata used for project creation.
get_projects0 of 3 parameters describedList/search projects with pagination.
create_project0 of 8 parameters describedCreate a new blank project in the current account organization and wait for it to appear in search. A project is a backend plus at most one client app, named by appType — pick it with the user: the free plan pays for one client. WEB, WECHAT_MINI_PROGRAM (Zion only) and MOBILE (iOS) each create a client this project hosts: build its pages with these tools, or deploy your own build to a WEB client with site deploy. BACKEND_ONLY creates no client, for a front end hosted elsewhere that calls the project's generated GraphQL API. These tools cannot add a client to an existing project, and a second one is a paid client slot bought in the console, so pick the client the user wants now. It comes back as builtInApp, already current, so UI calls need no appExId.
set_current_project0 of 2 parameters describedSet the project every later call defaults to. With no appExId it also makes that project’s only client current (the client session holds its pages and the project backend both), and a project with no client defaults to its backend document. A project with several clients is refused, listing them, and leaves the current project as it was: pages belong to one client, so pass appExId to name it — or appExId: null to address the backend alone.
get_current_projectno parametersGet the current project context: the pinned project and the client app calls default to (appExId null means they address the project’s own backend document).
rename_project3 of 4 parameters describedRename a project.
rename_project_app0 of 3 parameters describedRename an app inside a project.
delete_project0 of 3 parameters describedDelete a project. IRREVERSIBLE: requires a human to approve a native GUI dialog by retyping the project id; the agent cannot self-approve. Set MCP_HITL=off to fall back to the confirmProjectExId/force text gate in headless automation.
reset_project0 of 4 parameters describedReset a project. IRREVERSIBLE: requires a human to approve a native GUI dialog by retyping the project id; the agent cannot self-approve. Set MCP_HITL=off to fall back to the confirmProjectExId/force text gate in headless automation.
sync_backend3 of 6 parameters describedSave the schema and deploy/sync the project runtime backend. Returns ztype diagnostics and blocks on backend validation errors by default.
schema_statusno parametersShow daemon schema sessions and TTL state.
ztype_tool_calls3 of 7 parameters describedRun ztype Copilot tool calls against the loaded schema as one atomic batch: either every call lands or none does, and `applied` reports which. A batch ztype rejects, or whose patch trips validation, reports errors with errorCode/failedIndex and applies nothing — pass force to apply over a DEPENDENCY_CONFLICT. Use schema_undo to revert the last change.
schema_undo3 of 3 parameters describedUndo the most recent schema change made in this session by applying a reverse CRDT patch. Refuses if a concurrent edit touched the same path. Pairs with schema_redo.
schema_redo3 of 3 parameters describedRe-apply the schema change most recently undone via schema_undo in this session.
save_secret_value3 of 6 parameters describedPrompt the user (native GUI dialog) to set a secret's plaintext value, store it, and attach the handle so the secret reports hasValue:true. Provide secretId (an existing secret) OR displayName (create a new one). The value is typed into the dialog and never passes through this tool call or the model. Requires a desktop session (fails on headless/SSH/CI).
runtime_graphql3 of 5 parameters describedRun a raw GraphQL operation against the deployed project's runtime backend (the auto-generated data/actionflow/AI API), as admin.
runtime_run_code3 of 6 parameters describedExecute a JavaScript snippet in the deployed project's Run Code sandbox and return its result. Database writes are reverted unless updateDb is true, so this is the cheap way to check what a Run Code node will actually return before wiring it into a flow. Runs on the admin surface, as admin.
search_logs3 of 8 parameters describedFetch recent runtime logs from the project's log endpoint. Window is ±5 minutes around requestCreatedAt (ISO-8601, defaults to now). types: GATEWAY, ACTION_FLOW, ACTION_FLOW_NODE, ACTION_FLOW_CONTEXT_LOG, DEPLOYMENT, TPA, TRIGGER, SQL_GENERATION, GQL, ZAI. levels: INFO, WARNING, ERROR. customQueryCondition is an Elasticsearch condition (e.g. 'traceId: "abc-123"').
search_docs0 of 2 parameters describedFull-text search the bundled product documentation. Returns the best-matching pages as { path, title, url }, where url is the public HTTPS docs link you can cite. Use docs.get_page with a returned path to read a page.
get_doc_page0 of 1 parameter describedFetch a documentation page as markdown by its path (as returned by docs.search, e.g. "/02_design/01_layout_concepts").
platform_graphql0 of 2 parameters describedRaw GraphQL passthrough to the platform backend (projects, accounts, deployments — the editor/meta API, not the deployed runtime). Queries pass through unchanged; mutations are refused except a short allow list for client builds and code component versions, which the refusal names.
fetch_api_doc0 of 2 parameters describedFetch a web page or spec by URL for reading. Returns the page content (HTML converted to Markdown; long pages are windowed — pass `offset` with the same url to keep reading, guided by `totalLength`/`truncated`), same-site `pageLinks` to explore the documentation further, and any OpenAPI/Swagger `specLinks` found. Explore iteratively by fetching returned links, within a budget of 15 fetches per session (`fetchesRemaining`). Build the endpoints you read about with the api schema ops.
deploy_static_site3 of 9 parameters describedDeploy a built frontend directory (e.g. ./dist) to the project’s hosted static site and return the live url. Scans the directory, declares the manifest, uploads every asset to object storage, and switches the site over in one call — the agent never handles checksums or upload urls itself. `target` is BETA (default) or PROD; a PROD deploy takes the site live and requires a human to approve a native GUI dialog by retyping the project id; the agent cannot self-approve, and `force` does NOT approve it. Set MCP_HITL=off to fall back to the confirmProjectExId text gate in headless automation. An interrupted run resumes automatically when the directory is unchanged; pass `force` to discard it and start over. `dryRun` reports what would be uploaded without creating a deployment. The total size must fit the project’s remaining object storage, and the directory must contain an index.html.
get_static_site_status3 of 4 parameters describedReport the project’s static site: the currently live deployment (with its url) and any upload still in progress. Use to find the site url, or to see whether an interrupted deploy is still resumable.
abort_static_site_deployment3 of 5 parameters describedAbandon an in-progress static site upload and delete whatever it already uploaded. Targets the current in-progress deployment unless `deploymentExId` names one. The live site is untouched.
deploy_wechat_source_code3 of 10 parameters describedUpload a locally-developed WeChat mini program source tree (the directory holding app.json) to the project and start the server-side build that commits it to WeChat as the 体验版 (trial version). Scans the directory, declares the manifest, uploads every file to object storage, and hands off to the build job in one call — the agent never handles checksums or upload urls itself. By default it then WAITS for the build and returns the trial version number plus the path of the trial QR code image it saved (`wait: false` returns as soon as the build starts; `timeoutSeconds` bounds the wait). Requires the project to have granted Zion third-party platform authorization. An interrupted run resumes automatically when the directory is unchanged; pass `force` to discard it and start over. `dryRun` reports what would be uploaded without creating a deployment. This does NOT submit for audit or release — it only produces a trial version to verify; submitting it for review and releasing it are done by the user in the editor, not from here.
get_wechat_source_code_status3 of 3 parameters describedReport the project’s WeChat mini program source upload: any upload still in progress and any build already running (with its deploymentRecordExId and start time). Use to poll a build after `wechat deploy`, or to see whether an interrupted upload is still resumable.
abort_wechat_source_code_deployment3 of 4 parameters describedAbandon an in-progress WeChat mini program source upload and delete whatever it already uploaded. Targets the current in-progress deployment unless `deploymentExId` names one. A deployment that already reached BUILDING cannot be aborted.
get_wechat_trial_qrcode3 of 4 parameters describedSave the WeChat trial version (体验版) QR code of the project’s mini program to a local image file and return its path, along with the trial version number and the time it was committed. The QR code is written to disk rather than returned inline because it is a JPEG. Available once a `wechat deploy` build has succeeded; scanning it opens the trial version on a phone whose WeChat id is bound as a tester.
schema_session3 of 5 parameters describedSchema session lifecycle for the loaded project. op: LOAD (warm the in-memory CRDT session) | RELOAD (force reload from backend) | SYNC (catch up patches) | SNAPSHOT (return the data model: tables, relations, enum types; pass full=true for the entire raw document) | VALIDATE (ztype type checks, stable/beta/backend diagnostics). Edits do not go live until sync_backend.
schema_observe0 of 4 parameters describedRead the daemon's schema change-state for a project. op: STATE (revision, stale flag, latest change event) | EVENTS (recent semantic schema-change events; afterEventId/limit paginate).
auth0 of 1 parameter describedAuthentication. op: LOGIN (browser login + save credentials) | LOGOUT (delete credentials, unload sessions) | WHOAMI (show authenticated account token payload) | REAUTH (delete credentials and log in again).
project_inspect3 of 8 parameters describedInspect a project. op: DETAIL (app list, project config, latest schema metadata, deployment) | METADATA (plan/capability, deployment, app list, schema metadata, RPS summary) | RESOURCES (resource consumption, plan limits, RPS, optional history). includeSecrets applies to DETAIL/METADATA; refresh/resourceType/nonRealTimeDisplayedResourceTypes apply to RESOURCES.
actionflow3 of 5 parameters describedEdit server-side action flows — flows, nodes, branches, I/O, global variables, and triggers (scheduled / database-change). op is a schema ToolName; args is its payload. Call guide {plugin:"actionflow"} for the rules and arg shapes.
api3 of 5 parameters describedEdit the API-integration library — workspaces, APIs, request parameters, response configs, input variables, and workspace constants. op is a schema ToolName; args is its payload. Call guide {plugin:"api"} for the rules and arg shapes.
bindings3 of 5 parameters describedCreate and edit data bindings — option/const/formula/conditional bindings, boolean condition expressions, and request filters (where + sort), all addressed by schemaPath. op is a schema ToolName; args is its payload. Call guide {plugin:"bindings"} for the rules and arg shapes.
component3 of 5 parameters describedBuild and edit the UI — pages, containers, components, slots, properties, styles and responsive breakpoints, plus the design system (color themes, style scale), the client's app-wide variables, SEO path ranges, and custom verification-code delivery channels. op is a schema ToolName; args is its payload. Call guide {plugin:"component"} first for the tree rules, component templates, and per-op arg shapes.
database3 of 5 parameters describedEdit the data model — tables, columns, relations, unique constraints, display names. op is a schema ToolName from the enum; args is its payload. Call guide {plugin:"database"} first for the rules and per-op arg shapes.
payments3 of 5 parameters describedDesign and wire up payments: pick the model the project uses (legacy activation or payment configs), configure providers and credentials, and bind the callback fulfillment flows. op is a schema ToolName; args is its payload. Call guide {plugin:"payments"} for the rules and arg shapes.
permissions3 of 5 parameters describedEdit roles and their table, API, action-flow, AI, and payment permissions. op is a schema ToolName; args is its payload. Call guide {plugin:"permissions"} for the rules and arg shapes.
secret3 of 5 parameters describedRegister, update, and delete project secrets (API keys and tokens) by handle — never their plaintext. op is a schema ToolName; args is its payload. Call guide {plugin:"secret"} for the rules and arg shapes.
type3 of 5 parameters describedEdit enums and object/custom type definitions. op is a schema ToolName; args is its payload. Call guide {plugin:"type"} for the rules and arg shapes.
webhook3 of 5 parameters describedEdit incoming HTTP triggers (webhooks, called callbacks in the schema) — the bound action flow, request and response body shapes, body codecs, and the provider protocol. op is a schema ToolName; args is its payload. Call guide {plugin:"webhook"} for the rules and arg shapes.
zai3 of 5 parameters describedEdit ZAI AI-agent configs — agents, input args, and model settings. op is a schema ToolName; args is its payload. Call guide {plugin:"zai"} for the rules and arg shapes.
guide0 of 1 parameter describedLoad the domain guide for a backend capability — the rules, conventions, and gotchas for a plugin area (database, type, bindings, actionflow, api, zai, component, permissions, payments, logs, agent). Call before editing an unfamiliar area, or when a meta-tool call is rejected.
agent_command0 of 2 parameters describedAdvanced escape hatch for directly invoking an internal agent command.
Tested in a sandbox with an empty environment: no API keys, a throwaway folder, nothing from a real account. A passing test shows the server runs and responds. It is not a security audit.