What is Eve Horizon?
Eve Horizon is an agent-native development and deployment platform. Agents and humans share the same CLI and API with full parity — anything a human can do, an agent can do, and vice versa. Build, deploy, and operate any application, but especially applications where agents are first-class runtime components that autonomously develop, review, test, and ship software.
Think of it as the operating system for an automated software factory: requirements arrive from any channel — CLI, Slack, Nostr, GitHub, web chat — and agents bring them to life, with optional human review at any stage.
Core philosophy
Five ideas shape every design decision in the platform.
Agent-native. Agents are first-class citizens, not bolted-on afterthoughts. They authenticate with the same credentials, use the same CLI commands, and operate under the same permission model as human developers. The platform is designed so that agents can do everything — create projects, deploy services, run jobs, review code, and ship releases.
CLI-first. One interface for humans and agents alike. There is no hidden admin panel or separate agent API. The eve CLI is the universal entry point, whether you're typing commands or an agent is executing them programmatically.
Manifest-driven. Services, pipelines, environments, skills, and agent teams are all declared in a single manifest.yaml. One file describes your entire deployment shape — what gets built, how it flows through stages, and where it runs.
Event-driven. A central event spine in Postgres routes GitHub webhooks, Slack messages, cron ticks, and custom events into pipelines and workflows. Automation is declared in the manifest, not scattered across CI configs.
Skills-based. Reusable AI capabilities live in your repository as SKILL.md files following the OpenSkills format. Skills are installed from skill packs and read at runtime — no syncing, no copying, no drift.
Agents, packs, and the software factory
Eve agents receive work from anywhere — a Slack mention, a Nostr DM, a GitHub webhook, a web chat session, or a CLI command. The Chat Gateway normalizes all these channels through a pluggable provider architecture, routing messages to the right agent by slug or by pattern-matching rules.
Agents, skills, and teams are distributed through Agent Packs — composable bundles declared in your manifest:
x-eve:
packs:
- source: incept5/eve-skillpacks
ref: abc123...
- source: ./skillpacks/my-team
Packs define how agents work together. A pack might define a software factory — agents that take ideas from brief to production — or a review council, an incident response team, or any other knowledge-work pattern. Teams coordinate through dispatch modes (fanout, council, relay) with persistent coordination threads.
The software factory is one pattern, not a fixed pipeline. Each stage is an agent (or a team of agents) defined in a pack. By default agents work autonomously — human approval gates are opt-in, configured per environment or per skill when you want them.
What can you build?
| Use case | Description |
|---|---|
| Any application | Web apps, APIs, and microservices deployed to Kubernetes with managed pipelines, environments, and secrets. |
| Agentic-native apps | Applications where agents are runtime components — not just dev tools, but active participants in the running system. |
| Automated workflows | CI/CD pipelines, code review, documentation generation, incident response — all triggered by events and executed by agents. |
| Software factories | End-to-end autonomous development — agents take ideas from brief to production, with optional human review gates where you want them. |
Who is it for?
| Audience | What Eve gives you |
|---|---|
| Engineering teams | Deploy apps with managed pipelines and let agents handle review, testing, and documentation — defined in a single manifest file. |
| AI-forward organizations | Run autonomous agents at scale with governance, cost tracking, audit trails, and configurable approval gates. |
| Solo builders | A full platform in one CLI. Agents are your team — create a project, deploy to staging, and ship features without stitching together five SaaS tools. |
| Platform teams | Extensible primitives — harnesses, worker types, event triggers, agent teams — to build internal developer platforms with agent-native capabilities. |
Architecture at a glance
| Component | Role |
|---|---|
| API Gateway | REST API for all operations — projects, jobs, deployments, secrets, events. OpenAPI spec at /docs. |
| Gateway | Normalizes Slack, Nostr, chat, and webhook inputs for orchestration and job routing. |
| Event Spine | Central Postgres-backed bus that routes webhooks, messages, cron, and custom events into automation. |
| Orchestrator | Claims ready jobs, routes events to triggers, drives the job and deployment lifecycle. |
| Worker | Clones repos, spawns workloads, captures logs and artifacts in isolated spaces. |
| Agent Runtime | Executes agents and app-side orchestration logic for tasks and team workflows. |
| SSO | Federates authentication providers and token exchange paths for platform services. |
| Postgres | Single source of truth — jobs, events, secrets, logs, cost receipts. No hidden queues. |
| Kubernetes | Primary runtime. Each job attempt and service runs in isolated pods with dedicated workspaces. |
The platform runs on Kubernetes (primary) or Docker Compose (dev convenience).
Next steps
Ready to get started? Install the CLI and connect to the platform.