Self-host managed agents
Managed agents, inside your cloud.
Azents is an open-source control plane for remote agent runtimes.
Keep sessions, tools, permissions, and execution state next to your repos and private network.
remote runs
Close the laptop. The agent keeps working.
Durable runtimes keep sessions alive near the workspace.
laptop
Azents Runtime
workspace
Remote run basics
A run should survive disconnects, show progress, and resume without rebuilding local state.
Runs keep moving
Work continues after the tab, SSH session, or laptop is gone.
Workspace-native
Execute near the repo, files, shell, and task state.
Resume later
Pick up the same session without rehydrating your machine.
Visible progress
See tool calls, command output, diffs, and blockers.
Team-visible
A run belongs to the workspace, not one developer's laptop.
Right-sized runtime
Give heavier jobs the CPU, tools, and time they need.
Remote execution answers how agents keep working. Self-hosting answers where they run.
self-hosted runtime
Run managed agents as part of your cloud.
Keep runtimes close to repos, credentials, internal APIs, and audit trails.
Azents gives teams a self-hosted control plane for operating remote agent runs inside their own infrastructure.
how it works
One control plane. Your runtime boundary.
Azents keeps sessions in the control plane and runs workspace operations through runtimes in your infrastructure.
Engine
Distributed session and orchestration layer.
Runtime
Execution environment for code, files, and tools.
Session
Durable work context for long-running agent tasks.
Toolkit
Controlled access to private capabilities.
architecture
The parts that make remote agents durable.
Azents separates the session loop, runtime boundary, and provider-specific execution so agent work can survive real infrastructure.
Scroll, swipe, or use arrow keys. 01 / 03
self-bootstrapping
Azents is already using its own harness.
Azents is now used to develop Azents. The coding harness handles repositories, edits, shell, PRs, CI, review feedback, and verification.
$ azents run develop-azents
read the repository before changing it
edit files while preserving a working diff
run commands in controlled workspaces
open PRs and address review feedback
watch CI and isolate failures
iterate through implementation and verification
what you get
Managed-agent operations, self-hosted.
Not another chat UI. The operating layer for agents that work near real systems.
01
Managed sessions
Long-running work contexts with durable history.
02
Event history
Execution events for replay, audit, and inspection.
03
Engine / Runtime split
Scale orchestration separately from workspace execution.
04
Tool permissions
Control what an agent can read, run, call, and change.
05
Runtime providers
Run execution environments with Docker and Kubernetes-oriented providers.
06
Coding harness
Development workflows for repositories, PRs, and CI loops.
07
Self-hosted model
Designed for private cloud, on-prem, and regulated environments.
current focus
Built for teams operating agents close to real systems.
For platform engineers, infra teams, AI product builders, and organizations that need self-hosted control.
Focus areas
Self-hosted deployment docs
Engine and Runtime reliability
Runtime provider hardening
Coding harness maturity
Agent observability
Local LLM support
Broader agentic workflows
Self-host managed agents.
Run agent sessions, runtimes, and tools where your work already lives.