Multica
Multica coordinates coding agents through a shared workspace for issues, progress and deliverables. Teams assign work to an agent, inspect its runs and retain discussion alongside the result. The publisher calls it source-available and offers self-hosting, but execution still depends on the coding tools and runtimes connected to that workspace.
Keep orchestration separate from execution
The provider documentation says a runtime invokes the coding tool already installed and signed in on a computer. Multica does not replace that tool. Confirm the executable, account access and runtime status before expecting an assigned issue to produce work.
Model choices come from the runtime and can depend on local configuration or subscription entitlements. If no model is selected, the coding tool's default applies. Switching tools can make a previously selected model value invalid, so check it when changing an agent configuration.
Some tools support continuing an earlier session, provided the session still exists and the runtime can reach it. The docs say a new session is created when resumption fails. An issue's history therefore does not guarantee that every later run resumes the same local model context.
Review skills and connector configuration
Multica describes reusable skills bound to agents and injected through the coding tool's native discovery path. Existing repository skills are not overwritten on a name collision, according to the documentation. Review the injected instructions as part of reviewing what an agent is permitted to do.
Support for Multica-managed MCP differs by tool. That means a shared agent configuration is not proof that every selected runtime receives the same connectors. The docs also warn that MCP commands and environment values are saved server-side; do not include credentials you have no reason to share.
The workspace documentation lists scheduled runs and collaboration among agents. Test one bounded issue first and inspect its logs and resulting changes. Automatic scheduling does not establish that the delivered code is correct or ready to merge.
Choose the deployment and license deliberately
The about page says the source is public and self-hosting is available on your own infrastructure. It points separately to commercial licensing. Source availability should not be treated as an unrestricted license for every commercial deployment.
The retrieved pages do not establish a universal hosted price or include all external model costs. Review the workspace's applicable terms and the coding-provider access you intend to use.
Keep human review at the point where agent work becomes a production change. Multica's progress tracking helps make runs inspectable; it does not replace tests or an approval decision for consequential edits.