Project - AI management, online and open source.
A project management system where AI agents are first-class, not an integration.
Engineering teams ship alongside Codex, Claude Code, Cursor, Devin, hosted models, local models, and their own in-house agents. The work those agents do - the PRs they draft, the tasks they pick up, the reviews they run - doesn't fit cleanly into tools designed for humans passing tickets to humans.
Paimos treats agents as peers. Same board, same status, same audit trail as every other contributor. Built for teams doing agentic development, equally at home for the solo engineer running a kanban on their weekend project.
Code-aware agents get more than tickets: linked repos, a unified knowledge plane, canonical agent artifacts, issue → file anchors, and a mixed-context retrieval API /projects/:id/retrieve. They can also be handed work through explicit Codex and Claude Code runner actions, or asked for hosted/local draft output, with the result coming back to the same ticket history.
Two audiences, one tool - without the split personality.
§professional
Engineering teams doing agentic software development. Managing Codex and Claude Code runs, prompt presets, context packs, hosted drafts, local-model drafts, Cursor agents, internal frameworks - alongside humans - as part of one workflow.
Jira wasn't designed for this. Linear is closer but still treats AI as a sidecar. Paimos was built for the handoff, the context, and the audit trail from the name on.
§personal
Solo engineers with side projects. Same tool. No enterprise bloat, no process tax, no seat licenses.
A kanban for your weekend repo should feel as clean as one for a team of twenty - and not require you to pretend you're a "product owner".
Designed for you, the EU, and only god knows who!
Open source, agent-native, built in public. Phase 2 is now the shipping platform; Phase 3 preparation is visible in the product surface.
Paimos is under active development at github.com/markus-barta/paimos. v2.0 earned the Platform reading with structured agent context, a multi-action AI dispatcher, OpenAPI, and MCP. The v4.8 line hardens that platform: live Zitadel-backed SSO on the ppm reference deployment, knowledge entries outside ticket counts, explicit Codex and Claude Code Implement-this actions, profile/effort/prompt/context controls, repo-scoped local runners, hosted OpenRouter drafts, OpenAI-compatible local-model drafts, and the external Claude Code adapter path. Phase 1 (FOSS) stays active alongside; Phase 3 is about adoption readiness, not closing the codebase.
