Stoneforge is an orchestration platform for multi-agent coding workflows, with the core goal of enabling teams to run multiple AI coding agents in parallel and automatically handle merging and task distribution.
Purpose The official description points out that multi-agent collaboration commonly has three types of problems: merge conflicts (multiple agents editing the same files on the same branch), duplicate work (two agents grabbing the same task, or one task being blocked by another), and zero visibility (having to check terminals one by one to understand progress). Stoneforge is designed specifically to solve these problems.
Main capabilities
- Independent workspaces: each worker runs in an isolated git worktree without interfering with others; after committing and pushing, a merge steward is responsible for integration.
- Dependency-aware task distribution: the dispatch daemon tracks task priorities and dependencies, and automatically assigns ready tasks to idle workers, avoiding duplicate work and blocked startups.
- Real-time dashboard: you can view each agent's real-time output, kanban task progress, merge status, and performance metrics in a single Web dashboard.
- Real-time communication: channels, threads, and inbox triage allow agents to share context and report blockers.
- Evergreen documentation: a version-managed document library that supports full-text and semantic search to keep context up to date.
- Multi-provider support: compatible with Claude Code, OpenCode, and OpenAI Codex, and the official description says there is no lock-in.
- Linear-like issue tracking: supports priorities, dependencies, kanban views, and planning groups, all completed in the Web dashboard.
- Structured workflows: Playbook templates can be instantiated as resumable task sequences with persistent state.
- Automated merge review: the steward runs tests; if they pass, it squash-merges, and if they fail, it creates a repair task.
How to use The official description emphasizes "three-step command, get started in 30 seconds," but the page does not provide the specific command content. Users can visit the official website https://stoneforge.ai/ to get the actual installation and startup steps. Tasks can be assigned through the kanban, with example tasks such as "allow adjusting the task table column width by dragging and moving header borders," accompanied by priority values.
Applicable scenarios Suitable for development teams that need to have multiple AI coding agents process the same codebase in parallel and want to reduce merge conflicts and duplicate work, especially for scenarios where they want to monitor progress, merge status, and performance metrics in a unified dashboard.
Limitations The official description does not explain the specifically supported platforms, pricing, deployment methods, or scale limits in the provided content. The page mentions that "AI coding agents consume 20-40% of the context window on codebase exploration, and structured documentation can reduce location finding from 20 tool calls to 3," but it does not give complete benchmark details. Specific installation commands, system requirements, and pricing information are not yet provided.
