CloudBase AI ToolKit

Log in to CloudBase and bind an environment. Supports checking status, interactive login, API-key login, binding an environment with set_env, and logging out. An environment is the isolation unit for functions, databases and hosting — other tools act on it only once it is bound.

Call a Tencent Cloud or CloudBase control-plane API directly, by service, action and parameters. For integrating CloudBase business APIs over HTTP, the OpenAPI specifications are the better route; this is for control-plane actions the named tools do not cover.

Download a remote file into the project at a relative path you give — mini-program tab bar images and similar assets.

Download and deploy a CloudBase project template: React, Vue, WeChat mini-program, UniApp, or the editor rules alone for adding AI-editor configuration to a project you already have.

Add or remove an environment's security domains — the origins a browser client is allowed to call CloudBase from. Check the current list first, then add the exact host and port the browser sends. This is request-origin validation, not SSL certificates.

Read CloudBase environment information: the environment list, one environment's detail, or its security domains. An alias of queryEnv, kept for compatibility.

Create, update and delete remote agents.

Write application-side authentication configuration: login methods, providers, client configuration, publishable keys, API keys and custom login keys. It configures; it does not sign users in — client login belongs in your own code with the JS SDK.

Deploy a web application to CloudBase on its own subdomain: get a pre-signed upload URL, then upload the source and trigger the remote pipeline — install, build, deploy. Each service name gets its own subdomain.

Manage Cloud Run services in build order: initialise a project from a template, download service code, run it locally (function-type services), deploy, update configuration without redeploying code, and delete. A deploy onto an existing service merges with its remote configuration rather than replacing it.

Read data models: list them, read one model's schema, fields and relations, or generate SDK usage documentation for it. Read-only.

Manage environments: list available plans, create an environment, change its plan, or renew it. Everything here costs money and requires an explicit confirmation after the configuration summary is shown.

Write cloud functions: create them, update code and configuration, invoke them, manage timer triggers for scheduled jobs (seven-field cron), and bind layers. Container deployment creates an HTTP function from a TCR image.

Write the HTTP gateway: create, update and delete routes mapping a path to an upstream, and enable or disable existing routes. Without a domain it uses the default HTTP gateway domain, not the static hosting CDN domain.

Change static hosting: upload build output to the shared domain, delete files or directories with confirmation, set the index page, error page and routing rules, switch hosting on, bind, unbind or update custom domains, and download hosted content.

Manage CloudBase MySQL: provision it, destroy it, run write SQL and DDL, and initialise a schema. MySQL has to be provisioned before statements or schema initialisation will run.

Change permissions and users: resource permissions on databases, functions and buckets, roles, role members and policies, and application-user CRUD.

Write to a CloudBase PostgreSQL database: confirmed write SQL, a risk pre-check, and migrations. Schema changes — CREATE, ALTER, DROP — go through applyMigration with an explicit version, which writes or verifies a versioned migration file. Direct execute is for DML and operational SQL such as GRANT and RLS.

Manage cloud storage objects: upload, download and delete files and directories. Deletion requires force. Returned URLs are temporary signed links that expire in an hour — not addresses to persist. For PostgreSQL-mode environments use queryPgStorage instead.

Create a data model from a Mermaid class diagram. Creates new models only — it does not change an existing model's structure. Polls its own async task until it finishes.

Read remote agents: the list, one agent's detail, and its logs.

Read application-side authentication configuration: login methods, providers, publishable key, API keys, client configuration and static domains. It configures; it does not sign users in — client login belongs in your own code with the JS SDK.

Read deployed web applications: the application and version lists and their details. Poll a build by its id after deploying, and read the build log to diagnose a failure.

Read Cloud Run services: the service list, one service's detail, the available templates, and deployment logs. Service information includes name, status, access type, configuration and the most recent deployment context.

Read CloudBase environment information: the environment list, one environment's detail, or its security domains. The list action returns a summary only — id, alias, status, type, region and plan — with an exact-alias filter; full resource and billing detail requires reading one environment by id.

Read cloud function information: list functions and layers with paging, read one function's detail, its execution logs, its triggers, and its code download URL.

Read the HTTP gateway: routes under a domain and what each points at — cloud function, Cloud Run, static hosting or a lightweight server — plus custom domains, and the service and authentication switches. Domain plus path is the key.

Read static hosting: the site configuration — index page, error page, routing rules and site domains — hosting status, files by prefix, the full file list, and custom domain status. It changes nothing.

Check whether the CLS log service is switched on, and search logs across services. For one cloud function's own execution logs, use queryFunctions instead.

Read from CloudBase MySQL: read-only SQL, the provisioning result, task status and the instance's lifecycle context. Standard instance reads return no credentials; only the connection-info action passes the raw connection payload through, and that is for an explicit TCP migration.

Read permissions and users: resource permissions on databases, functions and buckets, the role list and role details, and application users. NoSQL collection rules and PostgreSQL row-level security are separate mechanisms.

Read from a CloudBase PostgreSQL database: the current context, database objects with their schema, lightweight metadata, one object's structure, and read-only SQL.

Read the storage capabilities of a PostgreSQL-mode environment: bucket and configuration summary, how to look up object information, and upload approaches over the HTTP API or the SDK. It reads no local files and does not emit signed URLs in bulk.

Read cloud storage: list a directory, read a file's information, and get a temporary signed download link whose lifetime you set. The public URL it derives is reachable only if the bucket is public-read; a default private bucket returns 403, so use the temporary URL. For PostgreSQL-mode environments use queryPgStorage instead.

Query records in a NoSQL collection, with filters, projection, sorting and paging.

Read NoSQL collection and index structure: list collections, read one collection's detail, list indexes and check whether an index exists. A management-side tool, not for writing client code.

Search CloudBase's knowledge base: fixed skill documents, OpenAPI specifications, official documentation, or a vector index. The fixed sources are the ones to reach for first; the vector search is for what they do not cover.

Insert, update, delete or upsert records in a NoSQL collection, with operators such as $set, $inc and $push. Server-side writes carry no _openid, so a collection whose security rules match on it needs that field supplied by hand or clients will not read the row back. Nested fields update by dotted path; passing a nested object replaces the whole field.

Create and delete NoSQL collections, and add or drop their indexes. A management-side tool, not for writing client code.