Oz Agent Worker
https://github.com/warpdotdev/oz-agent-worker- Category
- AI Agents
- Rank
- No. 1447Tools index
Previous survey · No. 1462 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- warpdotdev
- GitHub
- 42 stars
- Latest release
- v2026-09-07-20-38-22
- Date
About
Run Warp Oz cloud agents on your own infrastructure with full orchestration and observability.
What it does
Oz Agent Worker is a long-running task runner for Warp’s agent service. It receives assignments and cancellations over a WebSocket, then executes each task through Docker, directly on the host, as a Kubernetes Job, or by handing it to an operator command.
Why it's ranked here
Its strongest case is backend choice backed by concrete lifecycle behavior. Docker and Kubernetes enforce requested compute sizes, direct mode offers simple host execution, and command dispatch reaches custom runtimes. The worker also reconnects with backoff, handles cancellation, and reports classified failures. That breadth brings meaningful operational complexity.
What's good
Configuration supports command-line overrides and strict YAML validation. Kubernetes startup preflight catches incompatible sidecar loading before tasks arrive. Direct mode limits inherited host variables and isolates global Git configuration. Command dispatch sends task secrets through standard input instead of process arguments or the subprocess environment.
Tradeoffs
The service still depends on Warp authentication and outbound access to Warp servers. Recommended Docker deployment mounts the host socket, granting broad daemon access. Direct execution lacks container isolation and ignores compute shapes. Command-dispatched tasks outlive worker shutdown, do not occupy concurrency slots after dispatch, and must report their own completion.
How to use it well
It suits teams already running Oz tasks that need execution inside controlled Docker hosts, Kubernetes clusters, or custom schedulers. Pin an immutable container tag or digest, choose one backend, and set explicit concurrency and cleanup policy. Use Kubernetes when scheduling controls and resource enforcement matter. It does not replace Warp’s control service or provide a general-purpose agent platform.
Technical notes+
main.go defines the Kong CLI, merges flags over YAML configuration, initializes OpenTelemetry metrics, selects a backend, and handles graceful shutdown. internal/config/config.go enables strict known-field parsing, validates that at most one backend is configured, and resolves inherited environment entries. internal/worker/worker.go owns WebSocket reconnection, heartbeat, task tracking, cancellation, and optional semaphore limits. internal/worker/backend.go distinguishes completed, failed, and remotely spawned execution. internal/worker/docker.go restricts Docker operation to linux/amd64 and linux/arm64 and applies CPU and memory limits. internal/worker/direct.go creates protected workspaces, limits host environment inheritance, and invokes the local Oz CLI.
Observed
- Primary language
- Go, with module metadata declaring Go 1.26.5.
- Install surface
- Multi-architecture Docker images and a namespace-scoped Helm chart are documented.
- Interface
- Long-running CLI daemon using an authenticated WebSocket connection to receive tasks.
- Execution backends
- Docker, direct host execution, Kubernetes Jobs, and operator-defined command dispatch.
- Docker platforms
- Docker execution accepts linux/amd64 and linux/arm64 daemon platforms.
- Configuration
- Command-line flags and strict YAML configuration are supported.
Read from README.md, go.mod, main.go, internal/log/log.go, internal/config/config.go, internal/worker/direct.go, internal/worker/docker.go, internal/worker/errors.go, internal/worker/worker.go, internal/types/messages.go, internal/worker/backend.go.
What it can do
Deploy Warp Oz cloud agents
Agent configuration and target infrastructure details → Running agent instances on specified infrastructure
Orchestrate multiple cloud agents
Agent deployment specifications and workflow definitions → Coordinated execution across distributed agent instances
Monitor agent performance and health
Running agent instances and system metrics → Performance dashboards and health status reports
Track agent execution logs
Agent runtime activities and events → Detailed execution logs and audit trails
Scale agent infrastructure
Current resource utilization and scaling requirements → Adjusted infrastructure capacity and agent instances
Manage agent lifecycle operations
Agent deployment requests and operational commands → Started, stopped, updated, or terminated agent instances
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.