Apollo Deploy

Let your agents ship to production. You keep the approval.

Apollo Deploy is an open-source deployment platform for humans, CLIs, and AI agents. Every deploy starts as a plan you can read, so you can hand off more of the work and still know exactly what changed.

Built for the pace you ship at

Apollo Deploy is designed for the generative AI, hosting, payments, and streaming workloads that ship many times a day.

See the network
  • DigitalOcean
  • Meta
  • Microsoft
  • Signal
  • Sony Group
  • Vercel
  • X
  • Apple
  • Amazon
  • Netflix

From source code to a running application.
One loop, whoever is driving it.

Connect

Start from source. Connect a Git repository, upload an application, or deploy through the CLI or API.
  • Git repository
  • CLI and API
  • Runtime discovery

Plan

Know the change. Every caller produces the same deterministic plan—readable by a person, parseable by an agent—before resources are created, updated, or removed.
  • Readable changes
  • Policy checks
  • Machine-readable plan

Deploy

Apply safely. Build the application, provision its runtime, deploy the version, and route traffic with every action recorded.
  • Reproducible builds
  • Health verification
  • Traffic control

Monitor

Understand production. Watch health, errors, latency, resources, and deployment signals as one operating picture.
  • Runtime health
  • Release signals
  • Resource usage

Respond

Recover with control. Recommend or execute authorised scaling, traffic, restart, and rollback actions according to policy.
  • Recovery plans
  • Approval gates
  • Complete audit trail

Propose. Approve. Act.
The loop that lets you say yes.

SIGNALSrelease v42
p95 latency
137ms
error rate
0.03%
saturation
54%
LIVE60s window

Observe continuously.

Application and infrastructure signals stay connected to the release that produced them.

Approve explicitly.

Approval sits with a human, a policy, or a scope you set in advance—separate from whoever proposed the plan, so you can widen what agents do without losing the final say.

Audit completely.

Every plan, approval, action, and outcome stays legible long after the incident is closed.

Apollo Deploy

Access

How the agent reaches your environment.

Credentials it holds
None
Scoped to one project
Separate identity

Control

How a change gets from proposal to live.

Plan before it runs
Separate approval
Reversible in one step

Traceability

What you can see after the fact.

Action-level audit trail
Who changed what, and why
Agent with shell access

Access

How the agent reaches your environment.

Credentials it holds
Keys on the box
Scoped to one project
Root, in practice
Separate identity
Shares yours

Control

How a change gets from proposal to live.

Plan before it runs
No
Separate approval
No
Reversible in one step
Write a script

Traceability

What you can see after the fact.

Action-level audit trail
Shell history
Who changed what, and why
No
Agent with an API token

Access

How the agent reaches your environment.

Credentials it holds
A long-lived token
Scoped to one project
Only if you set it
Separate identity
Shares yours

Control

How a change gets from proposal to live.

Plan before it runs
No
Separate approval
No
Reversible in one step
Manual

Traceability

What you can see after the fact.

Action-level audit trail
Request logs
Who changed what, and why
Partial

One permission model. Four ways to drive it.

A person in the browser, a script in CI, a service on the API, or an agent in a loop. Same plan, same policy, same audit trail—because they are the same caller to us.

Read the documentation

Agents are not a plugin. They are one of the four callers this platform was designed around.

Web

Deploy and review plans with your team.

Command Line

Work locally, in scripts, and in CI pipelines.

API

Drive every operation programmatically.

AI Agents

Read state, propose plans, act within scope.

Your infrastructure or ours.

01
Your Own Infrastructure
02
Our Cloud
03
Hybrid
Run Apollo Deploy where it makes sense for your applications, data, and operational boundaries.
Keep control. Self-host on existing servers or private infrastructure.
Grow freely. Combine private workloads with Apollo Cloud capacity.

Infrastructure should never become another locked door.

Built in the open
Apollo Deploy is open source because the platform running your applications should be transparent, extensible, and yours to deploy without asking a vendor for permission. You can read exactly what an agent is allowed to do, because the policy engine is in the repository.
Shape what's next
Build deployment workflows, expose new capabilities to agents, self-host on your own hardware, and help shape what autonomous operations should look like. That is a better question to answer in the open.

Frequently asked questions

The practical details behind controlled application operations.

Can't find what you're looking for? Contact our support team

Apollo Deploy

Agents and automation

Can't find what you're looking for? Contact our support team

Let the agent propose. You approve.

Connect a Git repository, upload an application, or deploy through the CLI or API. Whoever starts the work—a person, a pipeline, or an agent—Apollo Deploy produces the same deterministic plan before resources are created, updated, or removed.
Approval is explicit, every action is recorded, and rollback is a first-class operation. Start with review on everything, then widen what runs unattended as the audit trail shows you what to trust.
Get early access