# Orca Workspace for parallel coding agents with separate Git worktrees, diff review, and local or self-managed runtimes. Canonical profile: https://ki-agentenkompass.de/en/agents/orca/ Editor: [Michael Meierhoff](https://ki-agentenkompass.de/en/about/) Sources reviewed: 2026-11-09 Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests. ## Overview Standalone workspace over existing coding agents; their models, accounts, and memory rules remain relevant. - Provider: Stably AI - Product family: Orca - Profile type: Open-source agent development environment - Environment: Desktop for macOS, Windows, and Linux; mobile companion plus SSH and your own Orca servers. - Connections: Claude Code, Codex, and other CLI agents; GitHub, Linear, browser, terminal, and plugins. - Setup: Install Orca, open a repository, configure a CLI agent with your own access, and create a worktree per task. - Pricing model: Orca is free under MIT. Model access and independently chosen cloud runtimes incur separate costs. ## Key features ### Parallel worktrees Separate checkouts for concurrent tasks. Source: https://github.com/stablyai/orca ### Chat and terminal Supported sessions can be used through either interface. Source: https://www.onorca.dev/docs/agents/native-chat ### Review changes Annotate diffs and return feedback to the agent. Source: https://github.com/stablyai/orca ### Your own runtimes Connect local machines, SSH hosts, and your own servers. Source: https://www.onorca.dev/docs/ways-to-run ### Mobile companion Receive agent notifications and reply on the go. Source: https://github.com/stablyai/orca ## Example workflows ### Fix bugs in parallel - Your input: Repository and independent bug reports. - Possible result: Separate changes in worktrees. - What to check: Review diffs and tests before merging. ### Continue on your server - Your input: Reachable Orca server and project access. - Possible result: Shared runtime for desktop and mobile control. - What to check: Check server availability and agent status. ## Requirements and limits ### Agent access Requirement: The chosen CLI agent still requires installation and authorization. Source: https://www.onorca.dev/docs/agents/supported ### Uneven integration depth Limitation: Structured chat supports selected agents; SSH and WSL use terminal-backed chat. Source: https://www.onorca.dev/docs/agents/native-chat ### Check approval mode Requirement: Agent documentation describes prefilled approval-bypass flags. Deliberately select Manual or appropriate custom arguments. Source: https://www.onorca.dev/docs/agents/supported ## A good fit if … - you coordinate multiple coding agents in parallel - you want to use your own machines and model access - you want to monitor and steer coding tasks on the go ## Less suitable if … - you expect a fully managed runtime - you need identical chat features for every agent Editorial assessment based on provider documentation. Not a product test. ## Usable AI models Fixed choice across providers Orca launches different CLI agents and offers model selection in supported chats. Universal model or BYOK support is not established. - Model families named: Anthropic (Claude), OpenAI (GPT), Google (Gemini) - Model choice: You choose - Your own API key: Not stated - Router or custom endpoint: Not stated - Local models: Not stated Documented · Sources accessed on: 2026-11-09 - [Orca · Agents](https://www.onorca.dev/docs/agents/supported) - [Orca · Chat](https://www.onorca.dev/docs/agents/native-chat) ## Europe & regional availability Product access, country availability and data processing are assessed separately. Reviewed scope: Orca · Documented product access Reviewed on: 2026-11-09 ### Access and regional conditions With requirements Downloads for macOS, Windows, and Linux are documented. Documentation supplies no country eligibility list; model providers have their own access conditions. Region covered: EU/EEA, United Kingdom and Switzerland Documented reason: The conditions of the specific access route and connected services apply. Official source: https://github.com/stablyai/orca ## Accounts & permissions Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use. Configuration covered: Orca · Account and permissions Sources reviewed: 2026-11-09 ### Account used Documented with conditions: Agents use the configured access and permissions of their host. Evidence: https://www.onorca.dev/docs/ways-to-run ### Admin approval Documented with conditions: Manual, Auto, and Yolo modes are available depending on the agent; existing sessions retain their mode. Evidence: https://www.onorca.dev/docs/agents/native-chat ### Data access & team sharing Documented: Files, Git, terminal, and browser on the selected local or self-managed remote host. Evidence: https://www.onorca.dev/docs/ways-to-run ## Prices and included usage Pricing checked: 2026-11-09 Orca is free under MIT. Model access and independently chosen cloud runtimes incur separate costs. ### Open Source Free; external usage billed separately MIT software; your own agent subscriptions and infrastructure. Price source: https://github.com/stablyai/orca Pricing details: https://github.com/stablyai/orca ## Operation & oversight Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured. Deployment covered: Orca · Execution and resumption Sources reviewed: 2026-11-09 ### Files and working state Documented: Hooks report working, waiting, and done. Evidence: https://www.onorca.dev/docs/agents/hooks-memory ### Sign-ins Documented with conditions: CLI agents use their configured accounts; capabilities vary by agent. Evidence: https://www.onorca.dev/docs/agents/supported ### Starting further runs Documented: Launch through the UI or CLI; scheduled automations are supported. Evidence: https://github.com/stablyai/orca ### Human intervention Documented with conditions: Stop, direct steering, and approval decisions are available in supported chat. Evidence: https://www.onorca.dev/docs/agents/native-chat ### Detecting errors Documented: The message queue pauses after failed turns. Evidence: https://www.onorca.dev/docs/agents/native-chat ### Resuming after failures Documented: With a live daemon, running processes reconnect; host reboots require restarting agents. Evidence: https://www.onorca.dev/docs/model/session-restore ### Editorial assessment Editorial assessment for the documented access route. Each criterion lists requirements and unresolved operating details. ## Memory and conversation continuity Scope: Orca session state and conventions of launched agents Sources accessed: 2026-11-09 Project instructions and session restoration are documented; durable learning follows the chosen agent. ### Storage and scope Partly documented: CLAUDE.md and AGENTS.md reside in the repository; Orca treats them as agent files. A separate store for learned memories is not established here. - [Orca · Memory](https://www.onorca.dev/docs/agents/hooks-memory) ### Selection and timing Partly documented: Repository hooks run at agent launch. Automatic selection of durable learned facts belongs to the agent and is unspecified here. - [Orca · Memory](https://www.onorca.dev/docs/agents/hooks-memory) ### Inspect, edit and delete Partly documented: Instruction files can be opened and edited in the explorer. This source does not describe separate learned-memory management. - [Orca · Memory](https://www.onorca.dev/docs/agents/hooks-memory) ### Agent and channel boundaries Partly documented: Instructions may live at repository root or in subdirectories; sessions belong to worktrees. Cross-agent learned memory is not established. - [Orca · Memory](https://www.onorca.dev/docs/agents/hooks-memory) - [Orca · Restore](https://www.onorca.dev/docs/model/session-restore) ### Retrieval Partly documented: Orca respects existing agent conventions. Its own semantic retrieval of learned memories is not publicly described. - [Orca · Memory](https://www.onorca.dev/docs/agents/hooks-memory) ### Session reset and compaction Partly documented: Layout and history restore; processes survive an app restart with a live daemon, but not a host reboot. Context compaction and memory reset depend on the agent and are unspecified here. - [Orca · Restore](https://www.onorca.dev/docs/model/session-restore) ## Context The chosen CLI agent still requires installation and authorization. Official product documentation reviewed. Workflows and assessments are editorial; hands-on product tests are outside the scope of this directory. ## Sources - https://www.onorca.dev/ - https://github.com/stablyai/orca - https://www.onorca.dev/docs/agents/supported - https://www.onorca.dev/docs/agents/native-chat - https://www.onorca.dev/docs/agents/hooks-memory - https://www.onorca.dev/docs/model/session-restore - https://www.onorca.dev/docs/ways-to-run ## Provider Stably AI develops Orca as an open development environment. - Operator: Stably AI - Organization: Software provider - Website: https://www.onorca.dev/ - Reviewed: 2026-11-09 - Provider source: https://www.onorca.dev/ [Sources and method](https://ki-agentenkompass.de/en/sources/)