|openfde run --repo --agent claude|codex |编排派遣：在任务自身的git工作树中生成一个编码代理，使用引用的上下文包，将其工作流入审计轨迹，然后从其退出标记（[DONE / BLOCKED / NEEDS_HUMAN]）继续执行任务;run ready 展开，run followup 继续同一会话并进行审核反馈 | |openfde worktree list/open/remove |执行者创建的任务树：分支、提交、脏文件;移除拒绝丢弃未合并的工作 | |openfde ontology |本体作为工件：每种类型的含义、每种类型的实例数、孤儿数、模式违背;--rdf 导出 RDF/OWL（类、关系、个体、带有来源的事实） | |openfde path |连接两个实体的最短引用事实链——CFO如何与对账工作流程相关 |

OpenFDE：面向FDE的AI工作空间

英语 |简体中文 |日本語 |西班牙语

以100倍速度交付AI解决方案。 面试变成记忆，记忆变成可追溯待办事项，待办事项变成编码代理工作——被评估限制。

OpenFDE 是一个面向前沿部署工程师的本地优先人工智能工作空间。

| openfde run --repo --agent claude|codex | Orchestrated dispatch: spawn a coding agent in the task’s own git worktree with the cited context pack, stream its work into the audit trail, and move the task on from its exit marker (DONE / BLOCKED / NEEDS_HUMAN); run ready fans out, run followup continues the same session with review feedback | | openfde worktree list/open/remove | Task worktrees the runner created: branch, commits ahead of base, dirty files; removal refuses to drop unmerged work | | openfde ontology | The ontology as an artifact: what each type means, instances per type, orphans, schema violations; --rdf exports RDF/OWL (classes, relations, individuals, facts with provenance) | | openfde path | Shortest chain of cited facts connecting two entities — how the CFO relates to the reconciliation workflow |

OpenFDE: AI workspace for FDEs

English | 简体中文 | 日本語 | Español

Deliver AI solutions 100x faster. Interviews become memory, memory becomes traceable todos, todos become coding-agent work — gated by evals.

OpenFDE is a local-first AI workspace for forward deployed engineers. Think of it as an enterprise brain built the ontology way: the customer organization — its goals, workflows, decisions, constraints, data sources, people and pains — captured as a typed, cited, time-aware graph that both humans and coding agents read from and write back to. It compiles engagement material — interviews, chat logs, documents, PDFs, images — into an ontology-backed operational memory, and closes the loop between humans and coding agents: agents pull tasks and context bundles from the ledger, execute, and write findings back — while the customer’s leadership watches progress live, with every claim citing its source.

Why

An FDE’s working state lives in three fragile places:

• Knowledge lives in conversations. Who trusts which data source, why a decision was made, which constraint blocks a workflow — said once in a meeting, lost two weeks later.
• Tasks live in heads. The gap between “heard it in an interview” and “dispatched it to a coding agent” has no system, no context, no trail back to the source.
• Verification lives in feelings. Agent output gets accepted by vibes instead of evals.

OpenFDE turns all three into one system, starting with memory.

What OpenFDE does

• Ontology-backed operational memory. A fixed FDE domain ontology — goals, workflows, decisions, constraints, data sources, pain points — constrains extraction, so what enters the ledger is operational knowledge, not prose. A dot-line-plane lens (value planes → business flows → decision points) organizes it the way leadership thinks.