Agent SDKs
Typed execute endpoint
Agent SDKs provide a typed execute endpoint with per-agent credentials. Each agent operates under its own identity and policy scope.
The governance layer that sits in front of the webhooks your AI agents call. Policy gates, approval workflows, atomic bundles, and OTEL traces — before any action executes.
Works alongside any orchestration runtime — Google ADK, LangChain, CrewAI, AutoGen, or custom. This is the governance layer, not the runtime.
Agent SDKs provide a typed execute endpoint with per-agent credentials. Each agent operates under its own identity and policy scope.
Enforce spend limits, throughput caps, and payload validation before any agent action executes. Non-conforming requests are blocked with a clear reason.
Group multiple agent actions into a single bundle. If any step fails, preceding steps are rolled back automatically.
Require human approval before sensitive actions proceed. Every execution is recorded as an OTEL trace for full auditability.
AgentDelivery ACP is the governance layer between your agent runtime and the systems it acts on. Bring your own orchestration framework — we govern the actions.
Google ADK, LangChain, CrewAI, AutoGen, Bedrock Agents, or custom
Policy gates · approvals · bundles · OTEL traces
Payments, CRM, internal APIs, partner integrations
There’s no separate “agent” tier to buy. Every AgentDelivery plan ships the full Agent Control Plane — policy gates, approvals, atomic bundles, and OTEL traces. Plans differ on volume, not safety.
| Plan | Monthly | Events / mo | Agent Control Plane |
|---|---|---|---|
| Free | €0 | 1,000 | ✓ Production (capped) |
| Starter | €19 | 25K | ✓ Included |
| Core | €49 | 100K | ✓ Included |
| Scale | €89 | 250K | ✓ Included |
| Enterprise | from €349 | 1M+ | ✓ Included |
Every plan runs the Agent Control Plane in production — the Free plan included, capped at 1,000 events/month with no credit card. Exercise SDKs, policy gates, and bundles end-to-end before you upgrade.
An Agent Control Plane is a governance layer that decides what actions an AI agent is allowed to execute. AgentDelivery’s ACP sits in front of the webhooks your agents call — enforcing policy gates (budget, rate, content), requiring human approval for sensitive actions, grouping multi-step operations into atomic bundles with rollback, and emitting OTEL traces for full auditability.
Google’s Agent Control Plane (part of Vertex AI / Google ADK) is an orchestration runtime — it runs and manages AI agents. AgentDelivery’s Agent Control Plane is a governance layer — it controls what webhook actions agents are allowed to trigger, regardless of which runtime runs them. The two are complementary: orchestrate with Google ADK, govern with AgentDelivery.
All of them. The ACP is included in every plan — Free (€0), Starter (€19/month), Core (€49/month), Scale (€89/month), and Enterprise. There’s no separate agent tier to buy: plans differ on volume (events, endpoints, retention), not on safety. The Free plan runs the same ACP in production, capped, with no credit card, so you can exercise it before upgrading.
Yes. AgentDelivery’s ACP is framework-agnostic. It sits in front of any webhook-driven action your agent takes, regardless of which orchestration framework runs the agent. Typed SDKs let agents call the execute endpoint directly with per-agent credentials and policy scope.
Yes. The Free plan runs the real ACP in production, capped at 1,000 events/month with no credit card — exercise SDKs, policy gates, and bundles end-to-end before upgrading. The ACP is included on every plan.
Start free in production — the real control plane, capped, no credit card — or jump straight to a paid plan. The control plane ships with every one.