Developer and Agent Hub

Connect AppDeploy before coding begins so your AI assistant can build with the deployment instructions, templates, and supported patterns it needs from the start.

Connect before you build

AppDeploy is most useful when it is available at the start of the build, before the AI assistant or coding agent chooses the architecture and writes the implementation. This lets the agent read AppDeploy’s deployment rules and receive supported scaffold or template guidance before design decisions become expensive to change.

Choose your path

Use the entry point that matches how you work.

AI assistant user: Connect your chat

Install AppDeploy in ChatGPT, Claude, or another supported assistant, authenticate, and start the build with AppDeploy connected.

Choose an installation guide

Coding agent developer: Connect through MCP

Add the production MCP endpoint to your coding agent, complete authentication, and load AppDeploy guidance before implementation.

Open the MCP quickstart

Agent or crawler: Read structured resources

Discover AppDeploy through Markdown, OpenAPI, OAuth metadata, the site agent index, and generated skill files.

Open the agent index

Quickstart

1. Connect AppDeploy

Choose your AI assistant or coding agent from the installation page, add AppDeploy, and complete user authentication.

2. Load deployment and SDK guidance before coding

Use every required input from the active tool schema. Call get_deploy_instructions before writing code; dual clients pass supports_direct_http_requests. If it accepts app_type and frontend_template, choose supported values and pass both to receive the correct scaffold. If no scaffold is returned and get_app_template is available, call it with every required input: supported app_type and frontend_template, a selected features list for agent and dual clients, and supports_direct_http_requests for dual clients. Before writing AppDeploy SDK code, call get_appdeploy_sdk_reference once per top-level SDK feature when available; otherwise use the SDK guidance returned with the template.

3. Build, test, and deploy

Have the agent define expected behavior, write tests, implement the app with the selected scaffold or template and SDK guidance, and call deploy_app.

4. Review the live result and iterate

Poll get_app_status, review deployment and QA feedback, open the returned live URL, and ask the agent to fix or improve the app as needed.

Capabilities and permissions

AppDeploy declares its tool permissions in OpenAPI and OAuth protected-resource metadata so clients can understand the access associated with each capability.

ScopeCapabilitiesUse
deploy:readRead deployment guidance; inspect apps, status, source, versions, logs, and QA results.Discovery and diagnosis
deploy:writeCreate or update deployments, upload assets, apply versions, and change app configuration.Build and iteration
deploy:deleteDelete an app after the user explicitly requests the destructive operation.Explicit deletion only

These scopes classify server-side permissions in a machine-readable form. Current MCP OAuth connections grant the full scope set after user authentication; independently reduced-scope grants require a separate authentication release. Some clients also hide mutation tools behind a client-side control, so follow the relevant installation guide and enable write tools where required.

Reference

MCP documentation

Connection instructions, authentication, workflow, tools, limits, and troubleshooting.

OpenAPI specification

OpenAPI 3.1 description of the MCP transport, OAuth discovery, errors, and permission scopes.

Agent content index

When-to-use guidance and links to AppDeploy’s agent-readable documentation.

Agent skills

Fallback integration files for supported agents that cannot connect through MCP.

Security

Platform controls, shared responsibility, data handling, and security reporting.

Changelog

Recent product, MCP, deployment, and platform changes.

Support

For setup and integration problems, start with MCP troubleshooting. If you still need help, email support@appdeploy.ai.