For agents
Agents connect over MCP, deploy with verification, and get repair context when apps fail.
Start here
claude mcp add --transport http fastdeploy https://platform.fastdeploy.app/api/mcp --header "Authorization: Bearer <token>"<token> is a bearer token from your platform login; treat it like a password — don't commit it or bake it into screenshots or config you share.
MCP reference
deploy_app | Create, deploy, and wait for healthy; returns logs on failure. |
|---|---|
update_app | Atomically redeploy code with verification. |
delete_app | Remove an app and its runtime resources. |
get_app | Get one app and its current state. |
list_apps | List apps available to the current account. |
validate_code | Run AST and security checks without deploying. |
create_function | Create a handle(event) function with an optional cron schedule. |
|---|---|
update_function | Update function code or its schedule. |
run_function | Run a function on demand. |
get_function_executions | Read recent function run results. |
get_logs | Read recent application logs. |
|---|---|
get_attention | List open incidents that need attention. |
get_fix_brief | Get code, evidence, and guidance for a failing app. |
sdk_docs | Check runner packages and platform SDK modules. |
From signal to working code
01
get_attention returns the open incident.
{
"app_id": "app-fn-7c2",
"kind": "function_failures",
"status": "open",
"consecutive_failures": 5,
"last_error": {
"type": "ModuleNotFoundError",
"traceback": "ModuleNotFoundError: No module named 'jsonn'"
}
}02
get_fix_brief collects code, evidence, and guidance.
{
"app_id": "app-fn-7c2",
"files": {"handler.py": "import jsonn\n..."},
"recent_failures": [
{"status": "error", "error": "No module named 'jsonn'"}
],
"guidance": [
"An import isn't available in the runner.",
"Check sdk_docs before changing dependencies."
]
}03
Call update_function with the corrected code.
{
"app_id": "app-fn-7c2",
"code": "import json\n\ndef handle(event):\n return {\"body\": json.dumps(event)}",
"schedule": "*/15 * * * *",
"verify": true,
"result": {
"status": "updated"
}
}04
The next run succeeds, and the incident resolves.
{
"app_id": "app-fn-7c2",
"incident": {
"kind": "function_failures",
"status": "resolved"
},
"executions": [
{"status": "success"}
]
}Failure thresholds
An incident opens after 3 consecutive failed scheduled runs or health checks. It resolves on the first success. Fix briefs never modify code — an agent (or you) does.
Same platform, another interface
fastdeploy attention
fastdeploy fix-brief <app-id>Connect once. Deploy, verify, and return to failures with the useful context already gathered.