Configuration Reference¶

Agent Gateway is configured via workspace/gateway.yaml, with environment variables taking precedence over file values.

Environment variable pattern: prefix AGENT_GATEWAY_, nested sections separated by __.

AGENT_GATEWAY_SERVER__PORT=9000
AGENT_GATEWAY_MODEL__DEFAULT=gpt-4o
AGENT_GATEWAY_PERSISTENCE__BACKEND=postgres

YAML ${VAR} interpolation: any value in gateway.yaml may reference environment variables using ${VAR_NAME}. Gateway raises an error at startup if a referenced variable is undefined.

persistence:
  url: ${DATABASE_URL}

Configuration is loaded once at startup. There is no live-reload of gateway.yaml.

GatewayConfig¶

Root configuration object. Loaded from workspace/gateway.yaml.

Field	Type	Default	Description
`timezone`	`str`	`"UTC"`	IANA timezone name used by the scheduler (e.g. `"America/New_York"`).
`server`	`ServerConfig`	—	HTTP server settings.
`model`	`ModelConfig`	—	LLM model settings.
`guardrails`	`GuardrailsConfig`	—	Execution safety limits.
`auth`	`AuthConfig`	—	Authentication settings.
`notifications`	`NotificationsConfig`	—	Notification backend settings.
`persistence`	`PersistenceConfig`	—	Database persistence settings.
`telemetry`	`TelemetryConfig`	—	OpenTelemetry settings.
`queue`	`QueueConfig`	—	Async execution queue settings.
`scheduler`	`SchedulerConfig`	—	Cron scheduler settings.
`context_retrieval`	`ContextRetrievalConfig`	—	RAG context retrieval settings.
`mcp`	`McpConfig`	—	MCP server connection settings.
`memory`	`MemoryConfig`	—	Agent memory settings.
`cors`	`CorsConfig`	—	CORS middleware settings.
`rate_limit`	`RateLimitConfig`	—	Rate limiting middleware settings.
`security`	`SecurityConfig`	—	Security headers middleware settings.
`proxy`	`ProxyConfig`	—	Trust upstream reverse-proxy forwarded headers.
`dashboard`	`DashboardConfig`	—	Built-in web dashboard settings.
`context`	`dict[str, Any]`	`{}`	Arbitrary key-value context injected into all agent prompts.

ServerConfig¶

Controls the embedded uvicorn server when using gw.run() or agents-gateway serve.

Env prefix: AGENT_GATEWAY_SERVER__

Field	Type	Default	Description
`host`	`str`	`"0.0.0.0"`	Bind address.
`port`	`int`	`8000`	Listen port.
`workers`	`int`	`1`	Number of uvicorn worker processes.

ModelConfig¶

Default LLM settings applied to all agents unless overridden in AGENT.md frontmatter.

Env prefix: AGENT_GATEWAY_MODEL__

Field	Type	Default	Description
`default`	`str`	`"gpt-4o-mini"`	Default model identifier (any LiteLLM model string).
`temperature`	`float`	`0.1`	Sampling temperature (0.0–2.0).
`max_tokens`	`int`	`50000`	Maximum output tokens per LLM call.
`fallback`	`str \\| None`	`None`	Model to fall back to on primary model failure.

GuardrailsConfig¶

Safety limits applied per execution. Triggers GuardrailTriggered when exceeded.

Env prefix: AGENT_GATEWAY_GUARDRAILS__

Field	Type	Default	Description
`max_tool_calls`	`int`	`20`	Maximum tool invocations per execution. Triggers reason `"max_tool_calls"`.
`max_iterations`	`int`	`10`	Maximum LLM reasoning steps per execution. Triggers reason `"max_iterations"`.
`timeout_ms`	`int`	`60000`	Maximum wall-clock time in milliseconds. Triggers reason `"timeout"`.
`max_delegation_depth`	`int`	`3`	Maximum depth for agent-to-agent delegation chains. Prevents infinite loops.

AuthConfig¶

Top-level authentication configuration. The fluent API (use_api_keys, use_oauth2) takes precedence over these settings.

Env prefix: AGENT_GATEWAY_AUTH__

Field	Type	Default	Description
`enabled`	`bool`	`True`	Enable authentication middleware.
`mode`	`str`	`"api_key"`	Auth strategy: `"api_key"`, `"oauth2"`, `"composite"`, `"custom"`, or `"none"`.
`api_keys`	`list[AuthKeyConfig]`	`[]`	API key records (used when `mode` is `"api_key"` or `"composite"`).
`oauth2`	`OAuth2Config \\| None`	`None`	OAuth2 configuration (used when `mode` is `"oauth2"` or `"composite"`).
`public_paths`	`list[str]`	`["/v1/health"]`	Paths exempt from authentication.

AuthKeyConfig¶

Field	Type	Default	Description
`name`	`str`	—	Human-readable name for the key (for logging/audit).
`key`	`str`	—	The plaintext API key. Hashed at load time; plaintext is not retained.
`scopes`	`list[str]`	`["*"]`	Scopes granted to this key. `"*"` grants all.

OAuth2Config¶

Field	Type	Default	Description
`issuer`	`str`	—	Token issuer URL (e.g. `"https://auth.example.com"`).
`audience`	`str`	—	Expected `aud` claim value.
`jwks_uri`	`str \\| None`	`None`	JWKS endpoint URL. Defaults to `{issuer}/.well-known/jwks.json`.
`algorithms`	`list[str]`	`["RS256", "ES256"]`	Allowed signing algorithms.
`scope_claim`	`str`	`"scope"`	JWT claim containing scopes. Use `"scp"` for Azure AD.
`clock_skew_seconds`	`int`	`30`	Allowed clock skew when validating `iat`/`exp`.

NotificationsConfig¶

Env prefix: AGENT_GATEWAY_NOTIFICATIONS__

Field	Type	Default	Description
`slack`	`SlackConfig`	—	Slack notification settings.
`webhooks`	`list[WebhookEndpointConfig]`	`[]`	Webhook endpoint definitions.
`webhook_secret`	`str`	`""`	Default HMAC-SHA256 signing secret for all webhooks.

SlackConfig¶

Field	Type	Default	Description
`enabled`	`bool`	`False`	Enable Slack notifications.
`bot_token`	`str`	`""`	Slack bot token (`xoxb-...`).
`default_channel`	`str`	`"#agent-alerts"`	Default channel for notifications.

WebhookEndpointConfig¶

Field	Type	Default	Description
`name`	`str`	—	Endpoint name (referenced by agents in `AGENT.md`).
`url`	`str`	—	HTTPS URL to POST notifications to.
`secret`	`str`	`""`	Per-endpoint HMAC-SHA256 signing secret. Overrides the global `webhook_secret`.
`events`	`list[str]`	`[]`	Event types to forward. Empty list means all events.
`payload_template`	`str \\| None`	`None`	Jinja2 template string for custom request body.

PersistenceConfig¶

Env prefix: AGENT_GATEWAY_PERSISTENCE__

Field	Type	Default	Description
`enabled`	`bool`	`True`	Enable persistence. When `False`, all repositories use null (in-memory) implementations.
`backend`	`str`	`"sqlite"`	Backend type: `"sqlite"` or `"postgres"`.
`url`	`str`	`"sqlite+aiosqlite:///agent_gateway.db"`	SQLAlchemy async DSN.
`table_prefix`	`str`	`""`	Optional prefix for all table names (e.g. `"ag_"`).
`db_schema`	`str \\| None`	`None`	PostgreSQL schema name. Auto-created on startup if it does not exist. Has no effect with SQLite.

The fluent API (use_sqlite, use_postgres) overrides these settings.

TelemetryConfig¶

Env prefix: AGENT_GATEWAY_TELEMETRY__

Field	Type	Default	Description
`enabled`	`bool`	`True`	Enable OpenTelemetry instrumentation.
`service_name`	`str`	`"agent-gateway"`	Service name reported in traces and metrics.
`exporter`	`str`	`"console"`	Exporter type: `"console"`, `"otlp"`, or `"none"`.
`endpoint`	`str`	`"http://localhost:4317"`	OTLP collector endpoint.
`protocol`	`str`	`"grpc"`	OTLP transport protocol: `"grpc"` or `"http"`.
`sample_rate`	`float`	`1.0`	Trace sampling rate (0.0–1.0).

QueueConfig¶

Env prefix: AGENT_GATEWAY_QUEUE__

Field	Type	Default	Description
`backend`	`str`	`"none"`	Queue backend: `"none"`, `"memory"`, `"redis"`, or `"rabbitmq"`.
`redis_url`	`str`	`"redis://localhost:6379/0"`	Redis connection URL.
`rabbitmq_url`	`str`	`"amqp://guest:guest@localhost:5672/"`	RabbitMQ AMQP URL.
`stream_key`	`str`	`"ag:executions"`	Redis Streams key for execution jobs.
`queue_name`	`str`	`"ag.executions"`	RabbitMQ durable queue name.
`consumer_group`	`str`	`"ag-workers"`	Redis Streams consumer group name.
`workers`	`int`	`4`	Number of concurrent worker coroutines per process.
`max_retries`	`int`	`3`	Maximum retry attempts for failed jobs.
`visibility_timeout_s`	`int`	`300`	Seconds a job is invisible to other workers while being processed.
`drain_timeout_s`	`int`	`30`	Seconds to wait for in-flight jobs to complete on shutdown.
`default_execution_mode`	`str`	`"sync"`	Default execution mode for new requests: `"sync"` or `"async"`.

SchedulerConfig¶

Env prefix: AGENT_GATEWAY_SCHEDULER__

Field	Type	Default	Description
`enabled`	`bool`	`True`	Enable the cron scheduler.
`misfire_grace_seconds`	`int`	`60`	Seconds past the scheduled time within which a missed fire is still executed.
`max_instances`	`int`	`1`	Maximum concurrent instances of the same schedule.
`coalesce`	`bool`	`True`	Collapse multiple missed fires into a single execution.
`distributed_lock`	`DistributedLockConfig`	—	Distributed lock settings for multi-instance deployments.

DistributedLockConfig¶

Prevents duplicate scheduled job firings when multiple gateway instances run concurrently. The instance that acquires the lock fires the job; all others skip that firing.

Env prefix: AGENT_GATEWAY_SCHEDULER__DISTRIBUTED_LOCK__

Field	Type	Default	Description
`enabled`	`bool`	`False`	Enable distributed locking. Set to `True` for multi-instance or multi-worker deployments.
`backend`	`str`	`"auto"`	Lock backend: `"auto"`, `"redis"`, `"postgres"`, or `"none"`. `"auto"` selects Redis when a Redis queue is configured, PostgreSQL when a PostgreSQL persistence backend is in use, and no-op otherwise.
`redis_url`	`str \\| None`	`None`	Redis connection URL. When `None`, falls back to `queue.redis_url`. Only used by the `redis` backend.
`key_prefix`	`str`	`"ag:sched-lock:"`	Prefix applied to all lock keys in Redis. Has no effect with the `postgres` backend.
`lock_ttl_seconds`	`int`	`300`	Lock expiry in seconds. Set this to a value greater than the maximum expected job duration to prevent a crashed instance from permanently holding a lock.

ContextRetrievalConfig¶

Controls RAG retriever behaviour during prompt assembly.

Env prefix: AGENT_GATEWAY_CONTEXT_RETRIEVAL__

Field	Type	Default	Description
`retriever_timeout_seconds`	`float`	`10.0`	Maximum time to wait for a retriever call before skipping it.
`max_retrieved_chars`	`int`	`50000`	Maximum characters injected from retriever results into the prompt.
`max_context_file_chars`	`int`	`100000`	Maximum characters read from workspace context files.

McpConfig¶

Controls timeouts for MCP server connections and tool calls. See the MCP Servers guide.

Env prefix: AGENT_GATEWAY_MCP__

Field	Type	Default	Description
`tool_call_timeout_ms`	`int`	`30000`	Maximum time in milliseconds to wait for a single MCP tool call to complete.
`connection_timeout_ms`	`int`	`10000`	Maximum time in milliseconds to wait for an MCP server connection to be established.

MemoryConfig¶

Global memory defaults. Individual agents override these in AGENT.md frontmatter.

Env prefix: AGENT_GATEWAY_MEMORY__

Field	Type	Default	Description
`enabled`	`bool`	`False`	Enable memory globally. Agents still need `memory.enabled: true` in their `AGENT.md`.
`max_injected_chars`	`int`	`4000`	Maximum characters of memory context injected into each prompt.
`extraction_model`	`str \\| None`	`None`	Model for memory extraction. Defaults to the primary model.
`auto_extract`	`bool`	`False`	Automatically extract memories from each conversation turn.
`max_memory_md_lines`	`int`	`200`	Maximum lines in a `MEMORY.md` file (file backend only).
`compaction`	`CompactionConfig`	—	Memory compaction settings.

CompactionConfig¶

Prevents unbounded memory growth by summarising or discarding old memories.

Env prefix: AGENT_GATEWAY_MEMORY__COMPACTION__

Field	Type	Default	Description
`enabled`	`bool`	`True`	Enable compaction.
`max_memories_per_scope`	`int`	`100`	Trigger compaction when this many memories exist for a scope.
`compact_ratio`	`float`	`0.5`	Fraction of memories to remove during compaction (oldest first).
`min_age_hours`	`int`	`24`	Never compact memories younger than this.
`importance_threshold`	`float`	`0.8`	Never compact memories with importance score at or above this value.
`decay_factor`	`float`	`0.95`	Relevance score multiplier applied per day since last access.

CorsConfig¶

Env prefix: AGENT_GATEWAY_CORS__

Field	Type	Default	Description
`enabled`	`bool`	`False`	Enable CORS middleware.
`allow_origins`	`list[str]`	`["*"]`	Allowed origin(s). Use explicit origins when `allow_credentials` is `True`.
`allow_methods`	`list[str]`	`["GET", "POST", "DELETE", "OPTIONS"]`	Allowed HTTP methods.
`allow_headers`	`list[str]`	`["Authorization", "Content-Type"]`	Allowed request headers.
`allow_credentials`	`bool`	`False`	Allow cookies / `Authorization` headers in cross-origin requests. Wildcard origin is rejected when `True`.
`max_age`	`int`	`3600`	Preflight cache duration in seconds.

RateLimitConfig¶

Env prefix: AGENT_GATEWAY_RATE_LIMIT__

Field	Type	Default	Description
`enabled`	`bool`	`False`	Enable rate limiting middleware. Requires `slowapi`.
`default_limit`	`str`	`"100/minute"`	Default rate limit applied to all endpoints. Uses slowapi rate string format (e.g. `"10/second"`, `"100/minute"`, `"1000/hour"`).
`storage_uri`	`str \\| None`	`None`	URI for shared rate limit storage (e.g. `"redis://localhost:6379"`). Required for consistent limits across multiple workers.
`trust_forwarded_for`	`bool`	`False`	Use the `X-Forwarded-For` header to identify clients. Enable when behind a reverse proxy.

SecurityConfig¶

Env prefix: AGENT_GATEWAY_SECURITY__

Field	Type	Default	Description
`enabled`	`bool`	`True`	Enable security headers middleware. Enabled by default (opt-out).
`x_content_type_options`	`str`	`"nosniff"`	Value for the `X-Content-Type-Options` header.
`x_frame_options`	`str`	`"DENY"`	Value for the `X-Frame-Options` header. Use `"SAMEORIGIN"` to allow same-origin framing.
`strict_transport_security`	`str`	`"max-age=31536000; includeSubDomains"`	Value for the `Strict-Transport-Security` header. Set to `""` to omit.
`content_security_policy`	`str`	`"default-src 'self'"`	Value for the `Content-Security-Policy` header on API paths.
`referrer_policy`	`str`	`"strict-origin-when-cross-origin"`	Value for the `Referrer-Policy` header.
`dashboard_content_security_policy`	`str`	(relaxed CSP)	CSP applied to `/dashboard` paths. Allows inline styles/scripts by default.

ProxyConfig¶

Trust X-Forwarded-* headers from an upstream TLS-terminating reverse proxy. See the mounting guide for background.

Env prefix: AGENT_GATEWAY_PROXY__

Field	Type	Default	Description
`trust_forwarded`	`bool`	`False`	Install Uvicorn's `ProxyHeadersMiddleware`. Turn on only when a trusted reverse proxy terminates TLS in front of the app.
`forwarded_allow_ips`	`str`	`"127.0.0.1"`	Comma-separated list of trusted peer IPs. Use `"*"` on platforms where the upstream path is segregated by a private network (Cloud Run, Fly.io, ALB).

Prefer the equivalent Uvicorn CLI flags in production:

uvicorn app:app --proxy-headers --forwarded-allow-ips='*'

The fluent gw.use_proxy_headers() method is a fallback for operators who cannot set Uvicorn flags (Gunicorn-with-uvicorn-worker, etc.).

DashboardConfig¶

Env prefix: AGENT_GATEWAY_DASHBOARD__

Field	Type	Default	Description
`enabled`	`bool`	`False`	Enable the dashboard at `/dashboard`. Opt-in.
`title`	`str`	`"Agent Gateway"`	Browser tab title and sidebar heading.
`subtitle`	`str`	`"AI Control Plane"`	Tagline displayed beneath the title in the sidebar and on the login page.
`logo_url`	`str \\| None`	`None`	URL of a branding image for the sidebar header and login page. When not set, the default Material hub icon is shown.
`favicon_url`	`str \\| None`	`None`	URL of a custom browser tab favicon.
`auth`	`DashboardAuthConfig`	—	Dashboard authentication settings.
`theme`	`DashboardThemeConfig`	—	Visual theme settings.

DashboardAuthConfig¶

Field	Type	Default	Description
`enabled`	`bool`	`True`	Require login to access the dashboard.
`username`	`str`	`"admin"`	Login username (password auth mode).
`password`	`str`	`""`	Login password. Empty string disables password protection (warned at startup).
`admin_username`	`str \\| None`	`None`	Separate admin account username. Admin users can toggle schedules and retry executions.
`admin_password`	`str \\| None`	`None`	Admin account password. Both `admin_username` and `admin_password` must be set to enable the admin account.
`login_button_text`	`str`	`"Sign in with SSO"`	Text on the SSO login button (OAuth2 mode).
`session_secret`	`str`	`""`	Secret for signing session cookies. Auto-generated if empty.
`oauth2`	`DashboardOAuth2Config \\| None`	`None`	OAuth2/OIDC configuration. Mutually exclusive with `password`.
`session_cookie_name`	`str`	`"agw_dashboard_session"`	Cookie name. Rename if you run multiple gateways on the same domain.
`session_cookie_same_site`	`"lax" \\| "strict" \\| "none"`	`"lax"`	`SameSite` attribute. `"none"` requires `https_only=True` per spec.
`session_cookie_https_only`	`bool \\| None`	`None`	Set the `Secure` attribute. `None` = auto: `True` when `proxy.trust_forwarded` is on, else `False`.
`session_cookie_domain`	`str \\| None`	`None`	Optional `Domain` attribute on the cookie.
`session_max_age_seconds`	`int`	`86400`	Session lifetime in seconds (default 24h).

DashboardOAuth2Config¶

Field	Type	Default	Description
`issuer`	`str`	—	OIDC issuer URL.
`client_id`	`str`	—	OAuth2 client ID.
`client_secret`	`str`	—	OAuth2 client secret (confidential client; required).
`scopes`	`list[str]`	`["openid", "profile", "email"]`	OAuth2 scopes to request.

DashboardThemeConfig¶

Field	Type	Default	Description
`mode`	`"light" \\| "dark" \\| "auto"`	`"auto"`	Color scheme. `"auto"` follows the user's OS preference.
`accent_color`	`str`	`"#6366f1"`	Legacy accent color. Maps to `colors.primary` if `colors.primary` is unchanged.
`accent_color_dark`	`str`	`"#818cf8"`	Legacy dark accent. Maps to `colors.primary_dark` if unchanged.
`colors`	`DashboardColorConfig`	—	Granular color overrides.

DashboardColorConfig¶

All values are CSS hex color strings.

Field	Default (light)	Default (dark)	Description
`primary`	`"#6366f1"`	`"#818cf8"`	Buttons, links, active nav items.
`secondary`	`"#64748b"`	`"#94a3b8"`	Secondary actions, muted text.
`accent`	(= primary)	(= primary_dark)	Highlight color. Defaults to `primary` if empty.
`surface`	`"#ffffff"`	`"#141b2d"`	Card and panel backgrounds.
`sidebar`	`"#0f172a"`	`"#0b0f1a"`	Sidebar background.
`danger`	`"#ef4444"`	`"#f87171"`	Error states and destructive action buttons.

Full gateway.yaml Example¶

timezone: Europe/London

server:
  port: 9000
  workers: 4

model:
  default: gpt-4o
  temperature: 0.2
  max_tokens: 8192

guardrails:
  max_tool_calls: 30
  timeout_ms: 120000

auth:
  enabled: true
  mode: api_key
  api_keys:
    - name: my-service
      key: ${API_KEY}
      scopes: ["*"]

persistence:
  backend: postgres
  url: ${DATABASE_URL}

telemetry:
  exporter: otlp
  endpoint: http://otel-collector:4317

queue:
  backend: redis
  redis_url: ${REDIS_URL}

mcp:
  tool_call_timeout_ms: 30000
  connection_timeout_ms: 10000

notifications:
  slack:
    enabled: true
    bot_token: ${SLACK_BOT_TOKEN}
    default_channel: "#ops-alerts"

cors:
  enabled: true
  allow_origins:
    - https://app.example.com
  allow_credentials: true

dashboard:
  enabled: true
  title: My Agents
  auth:
    password: ${DASHBOARD_PASSWORD}
  theme:
    mode: dark

AGENT.md Frontmatter Fields¶

Each agent is defined by an AGENT.md file with YAML frontmatter. The following fields are recognized:

Field	Type	Default	Description
`description`	`str`	`""`	Human-readable agent description.
`display_name`	`str \\| null`	`null`	Optional display name (defaults to directory name).
`tags`	`list[str]`	`[]`	Tags for filtering and categorization.
`version`	`str \\| null`	`null`	Agent version string.
`enabled`	`bool`	`true`	Whether the agent is enabled. Disabled agents return 422 from invoke/chat. Editable from the admin dashboard.
`skills`	`list[str]`	`[]`	Skill IDs available to the agent.
`model`	`object`	`{}`	Model configuration (`name`, `temperature`, `max_tokens`, `fallback`).
`execution_mode`	`"sync" \\| "async"`	`"sync"`	Default execution mode.
`schedules`	`list[ScheduleConfig]`	`[]`	Cron schedule definitions. See ScheduleConfig below.
`scope`	`"global" \\| "personal"`	`"global"`	Agent scope. Personal agents require per-user setup.
`delegates_to`	`list[str]`	`[]`	Optional allow-list of agent IDs this agent can delegate to. When empty (default), the agent can delegate to any enabled agent.
`input_schema`	`object \\| null`	`null`	JSON Schema for input validation.
`setup_schema`	`object \\| null`	`null`	JSON Schema for personal agent setup form.
`notifications`	`object`	`{}`	Notification targets for `on_complete`, `on_error`, `on_timeout`.
`context`	`list[str]`	`[]`	Explicit context file paths relative to workspace root.
`retrievers`	`list[str]`	`[]`	Context retriever IDs for RAG.
`mcp_servers`	`list[str]`	`[]`	MCP server names this agent can access. If no agent lists `mcp_servers`, all agents can use all MCP tools.
`memory`	`object`	`{}`	Memory configuration (`enabled`, `auto_extract`, etc.).

ScheduleConfig¶

Each entry in the schedules list supports the following fields:

Field	Type	Default	Description
`name`	`str`	—	Unique schedule name within the agent. Forms the schedule ID as `{agent_id}:{name}`.
`cron`	`str`	—	Standard 5-field cron expression (e.g. `"0 9 * * 1-5"`).
`message`	`str`	—	Message sent to the agent when the schedule fires.
`input`	`dict`	`{}`	Additional structured input passed alongside the message.
`enabled`	`bool`	`true`	Whether the schedule is active. Set to `false` to define without activating.
`timezone`	`str`	`"UTC"`	IANA timezone name for the cron expression (e.g. `"America/New_York"`).
`instructions`	`str \\| None`	`None`	Per-schedule instructions injected into the agent's system prompt when this schedule fires. Allows a single agent to behave differently across multiple schedules (e.g. different post styles per schedule). Instructions appear after the agent's base prompt and any user personalization.

See the Scheduling guide for usage examples and guidance on writing effective per-schedule instructions.

Schedule `source` field¶

Every schedule record — whether loaded from AGENT.md or created at runtime via the API or dashboard — carries a source field that identifies its origin.

Value	Meaning
`"workspace"`	Defined in `AGENT.md` frontmatter. Re-synced on every workspace reload. Cannot be deleted via the API.
`"admin"`	Created at runtime by an admin user via `POST /v1/schedules` or `gw.create_admin_schedule()`. Persisted in the database. Survives workspace reloads and gateway restarts. Can be deleted via `DELETE /v1/schedules/{id}` or `gw.delete_admin_schedule()`.

The source field is included in all GET /v1/schedules and GET /v1/schedules/{id} responses. Workspace schedules cannot be deleted through the API or dashboard; the API returns 400 if a delete is attempted on a "workspace" schedule.

See the Admin-Created Schedules guide for full usage details.