For the complete documentation index, see llms.txt.
Skip to main content
Version: 8.10 (unreleased)

System requirements

Learn about the system requirements for running ProcessOS Harness.

Requirements​

ProcessOS Harness requires the following:

System overview: the builder client hosts the AI coding agent and a local Camunda server, syncs status with the ProcessOS governance process on the enterprise Camunda server, uses VCS and an AI platform, and generates a ProcessOS solution deployed back to the enterprise Camunda server.

note

Agentic tooling is flexible by nature, so ProcessOS Harness may run on more systems than the ones listed here. These are the configurations Camunda currently tests, and the goal is to keep ProcessOS Harness independent of any specific AI coding agent or AI platform.

Governance process​

The governance process needs a Camunda cluster that's reachable across your organization, because it holds ProcessOS Harness project state and orchestrates SME feedback.

RequirementSupported
Camunda 8A supported release of Camunda 8 SaaS or Self-Managed.

Project version control​

ProcessOS Harness projects are stored in repositories, so managing one requires a Git-compatible version control system. You can host it internally or use a provider such as GitHub.

AI coding agent​

ProcessOS Harness drives an AI coding agent to do the work of re-engineering.

AspectSupported
AI coding agentClaude Code and GitHub Copilot CLI. Other agents may be supported in future releases, but aren't yet tested exhaustively.
AI platformPlatforms such as Amazon Bedrock, Azure OpenAI, and Ollama, where the coding agent supports them.
ModelsModels natively used by a supported coding agent. Anthropic Claude models are tested the most, with Opus or smarter alternatives recommended.

The agent needs these permissions in your environment:

  • Outbound network access, for example curl and wget.
  • Write access within the ProcessOS Harness project Git repository.
  • Permission to execute bash commands and scripts.

AI coding agents are read-only by default and ask for explicit approval otherwise. Teams typically allowlist frequently used safe commands, or enable an accept-edits mode, to reduce the number of prompts during a run.

ProcessOS Bundle mapping​

ProcessOS Harness ships one bundle per AI coding agent. Every bundle carries the same skills, rules, and hooks, generated into the layout that agent expects. Install the bundle that matches your agent, because an agent only discovers skills in its own directory. To understand how bundles are installed, see Install ProcessOS Harness.

AI coding agentBundleInstall commandSkills directory
Claude Code (via CLI)claudecodec8 os install claudecode.claude/skills/
Claude Desktopclaudecodec8 os install claudecode.claude/skills/
GitHub Copilot CLIcopilotclic8 os install copilotcli.github/skills/

Only one bundle can be installed at a time. To move to a different agent, run c8 os switch <bundle>, which removes the current bundle before installing the new one.

Builder client​

The AI coding agent runs on the builder's own computer.

RequirementDetails
c8ctlThe Camunda 8 CLI, used to deploy and manage Camunda resources and clusters.
Operating systemWindows (via WSL), macOS, or Unix.
AI coding agentInstalled locally.
CamundaA local cluster started with c8run, used to validate generated solutions.
Node.jsVersion 22.18.0 or later.
Git and GitHub CLIgit, and the gh CLI for the GitHub discovery specialist.
Maven (For Java Workers only)Latest stable release.
Camunda ModelerFor editing BPMN and DMN files during review.

Generated solutions​

Solutions generated by ProcessOS Harness use the following technologies.

ComponentTechnology
WorkersTarget SDKs: Spring Boot; TypeScript; Python
Camunda artifacts.bpmn, .dmn, and .form files for the latest Camunda 8 version.
AI platformsAs supported by agentic orchestration.

TypeScript and Python are also fully supported worker targets, described in generate Camunda artifacts.