Turn tickets into merged pull requests
All your agent work. Task: All your agent work. One place to run it. Self-hosted orchestration for coding agents, automated workflows, and interactive sessions.On your Kubernetes cluster. On your own machines. In your pocket. Product tour · Get started · Documentation · Releases Know what is running, what is ready, and what needs you. All screenshots use fictional demonstration data; activity and usage are simulated. Optio gives your agents a shared workspace. Hand off a code change, schedule a report, respond to an event, or keep a terminal open while you move between your desk and your phone. Start and follow all of it from one Work feed. Choose Claude Code, OpenAI Codex, GitHub Copilot, Google Gemini, Cursor, OpenCode, Pi, or OpenClaw—or run a plain shell command. Pick the runtime for each piece of work, with the models and providers it supports. What can you do with Optio? The fourteen trigger types are manual, schedule, webhook, ticket, GitHub, GitLab, Slack, Linear, Jira, Pylon, PagerDuty, Sentry, Alertmanager (Prometheus / Grafana), and Datadog. Persistent agents can also wake on messages. GitHub, GitLab, and AWS CodeCommit are supported repository platforms; available CI and issue features depend on the platform. Repo work, jobs, local automations, and persistent agents share the same feed. Five answers describe the work The form derives the right kind of work from these choices and reads it back as a sentence. Edit recurring work in the same form. Choose a trigger, then shape the environment: pod or machine, repository and branch, ownership, connections and MCP tools, skills, setup commands, and PR review settings. Follow the repo's defaults or customize this piece of work. Stay close to your sessions Read the agent’s conversation or use its terminal. Open a shell in the same directory with Terminal here, split sessions side by side, and drag the session sidebar to the width you want. Pane arrangements stay on the current device. Session sharing requires organization sign-in and grants view and control. Links expire and can be revoked. Collaborators gain access to the session’s environment and capabilities, including credentials available there. Sharing and recovery details → Native apps, wherever you are Optio has a SwiftUI iOS app and a Jetpack Compose Android app. Check the queue, create work, read conversations, control terminals, and manage your workspace from your phone. iOS: home-screen and Lock Screen widgets, Live Activities, Dynamic Island, and a mirrored Live Activity in the Apple Watch Smart Stack. Android: home-screen widgets, notifications, and an ongoing watch notification. Contextual actions take you back to the work that needs you. Native Apple component captures with sample states. Watch support is provided by the iPhone Live Activity; there is no separate Watch app. Build iOS · Build Android · More screenshots Your infrastructure, your choices Cluster and laptop execution. Repository work uses worktrees in Kubernetes; Optio Local pairs your machines through an outbound connection and uses their CLI configuration. Connections and tools. Combine encrypted credentials, MCP servers, shell environment, and usage notes. Assign access by repo and runtime, with per-work settings. Connections → Explicit ownership and isolation. Execution pools include workspace, owner, purpose, and access settings. Work inside the same pool remains mutually trusted; ordinary pods are not a boundary for hostile tenants. Security review → Recovery with visible status. Sessions report reconnecting, resumable, or lost states. Preserve recoverable work and require an explicit decision before retrying an uncertain outcome. Recovery → Managed Kubernetes deployment. Helm supports managed PostgreSQL and Redis, existing Kubernetes Secrets for database, Redis, encryption, and OAuth settings, and separate worker identities. The combined API/web deployment runs any number of replicas (api.replicas or api.autoscaling) on one Postgres and one Redis; worker pods scale independently. EKS and production guide → Identity and access. Workspaces, admin/member/viewer roles, OAuth/OIDC sign-in, personal access tokens, and encrypted secrets at rest. Sign-in → Optio is the orchestration layer you host. Agents still communicate with the model providers and external tools you configure; self-hosting Optio does not make those services local. Get started For a local deployment, install Docker Desktop and enable Kubernetes, then:; tools: Claude Code, Codex. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include a Kubernetes cluster or local Docker environment, plus valid API keys for model providers (e.g., Anthropic, OpenAI) and repository access. Self-hosting Optio does not make model inference local; agents still communicate with external providers, incurring their costs and latency. The source notes that ordinary pods are not a security boundary for hostile tenants, requiring careful isolation configuration. Human approval is implied by the 'wait for you' exit option but not mandated by default..
Tools
Published on Agent Work Atlas:
Overview
All your agent work.
Workflow
All your agent work. One place to run it. Self-hosted orchestration for coding agents, automated workflows, and interactive sessions.On your Kubernetes cluster. On your own machines. In your pocket. Product tour · Get started · Documentation · Releases Know what is running, what is ready, and what needs you. All screenshots use fictional demonstration data; activity and usage are simulated. Optio gives your agents a shared workspace. Hand off a code change, schedule a report, respond to an event, or keep a terminal open while you move between your desk and your phone. Start and follow all of it from one Work feed. Choose Claude Code, OpenAI Codex, GitHub Copilot, Google Gemini, Cursor, OpenCode, Pi, or OpenClaw—or run a plain shell command. Pick the runtime for each piece of work, with the models and providers it supports. What can you do with Optio? The fourteen trigger types are manual, schedule, webhook, ticket, GitHub, GitLab, Slack, Linear, Jira, Pylon, PagerDuty, Sentry, Alertmanager (Prometheus / Grafana), and Datadog. Persistent agents can also wake on messages. GitHub, GitLab, and AWS CodeCommit are supported repository platforms; available CI and issue features depend on the platform. Repo work, jobs, local automations, and persistent agents share the same feed. Five answers describe the work The form derives the right kind of work from these choices and reads it back as a sentence. Edit recurring work in the same form. Choose a trigger, then shape the environment: pod or machine, repository and branch, ownership, connections and MCP tools, skills, setup commands, and PR review settings. Follow the repo's defaults or customize this piece of work. Stay close to your sessions Read the agent’s conversation or use its terminal. Open a shell in the same directory with Terminal here, split sessions side by side, and drag the session sidebar to the width you want. Pane arrangements stay on the current device. Session sharing requires organization sign-in and grants view and control. Links expire and can be revoked. Collaborators gain access to the session’s environment and capabilities, including credentials available there. Sharing and recovery details → Native apps, wherever you are Optio has a SwiftUI iOS app and a Jetpack Compose Android app. Check the queue, create work, read conversations, control terminals, and manage your workspace from your phone. iOS: home-screen and Lock Screen widgets, Live Activities, Dynamic Island, and a mirrored Live Activity in the Apple Watch Smart Stack. Android: home-screen widgets, notifications, and an ongoing watch notification. Contextual actions take you back to the work that needs you. Native Apple component captures with sample states. Watch support is provided by the iPhone Live Activity; there is no separate Watch app. Build iOS · Build Android · More screenshots Your infrastructure, your choices Cluster and laptop execution. Repository work uses worktrees in Kubernetes; Optio Local pairs your machines through an outbound connection and uses their CLI configuration. Connections and tools. Combine encrypted credentials, MCP servers, shell environment, and usage notes. Assign access by repo and runtime, with per-work settings. Connections → Explicit ownership and isolation. Execution pools include workspace, owner, purpose, and access settings. Work inside the same pool remains mutually trusted; ordinary pods are not a boundary for hostile tenants. Security review → Recovery with visible status. Sessions report reconnecting, resumable, or lost states. Preserve recoverable work and require an explicit decision before retrying an uncertain outcome. Recovery → Managed Kubernetes deployment. Helm supports managed PostgreSQL and Redis, existing Kubernetes Secrets for database, Redis, encryption, and OAuth settings, and separate worker identities. The combined API/web deployment runs any number of replicas (api.replicas or api.autoscaling) on one Postgres and one Redis; worker pods scale independently. EKS and production guide → Identity and access. Workspaces, admin/member/viewer roles, OAuth/OIDC sign-in, personal access tokens, and encrypted secrets at rest. Sign-in → Optio is the orchestration layer you host. Agents still communicate with the model providers and external tools you configure; self-hosting Optio does not make those services local. Get started For a local deployment, install Docker Desktop and enable Kubernetes, then:
Reported result
The source provides no measured outcomes, success rates, or specific business impact for this workflow. It explicitly states that all screenshots use fictional demonstration data and simulated activity. Therefore, actual performance metrics are unknown. A proposed acceptance check would be to measure the time-to-merge for a sample of 10 tickets compared to manual baseline, but this data is not present in the text.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include a Kubernetes cluster or local Docker environment, plus valid API keys for model providers (e.g., Anthropic, OpenAI) and repository access. Self-hosting Optio does not make model inference local; agents still communicate with external providers, incurring their costs and latency. The source notes that ordinary pods are not a security boundary for hostile tenants, requiring careful isolation configuration. Human approval is implied by the 'wait for you' exit option but not mandated by default.
Try it
I want to set up a ticket-to-PR workflow using Optio. First, deploy Optio locally using Docker Desktop and Kubernetes, then configure a connection to my [replaceable: GitHub repository URL] and a model provider like [replaceable: Claude Code or Codex]. Create a new work item triggered by [replaceable: Jira or GitHub Issue] events. Configure the agent to work in a repo worktree, open a PR, and pause for human review before merging. Run a dry test with a single low-risk ticket [replaceable: ticket ID] to verify the PR creation and CI feedback loop. Do not auto-merge. Show me the agent's conversation log and the resulting PR diff. If the output is correct, I will approve the merge manually. If the agent fails to open a PR or ignores CI errors, stop and report the error logs. Ensure no sensitive credentials are exposed in the logs.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.