Plug in the systems you already run. No rip-and-replace.
Wrap the systems you already run as tools an agent can call, without rebuilding them.
Plug in what you already run. Sequence the work. Pause for a person on writes.
Wrap the systems you already run as tools an agent can call, without rebuilding them.
Tasks are decomposed, run in parallel where dependencies allow, and stay inspectable.
Write operations pause for confirmation, and every operation leaves a record.
The community edition is the full feature set. Docker Compose brings it up on your own box in under five minutes, and the data stays on your network.
git clone https://github.com/fim-ai/fim-one.git
cd fim-one && cp example.env .env && ./start.sh
# → http://localhost:3000Source-available · Self-hosted · Data never leaves your network
Contracts and compliance files sat in different systems. FIM One connected them, so one question returns cited answers.
DingTalk, WeCom, and Feishu were siloed. One FIM One agent spans all three, self-hosted, data staying on the network.
We needed case and regulatory knowledge searchable across systems, running entirely on our own network. Few platforms deploy on-premise and still connect everything.
Most AI tools assume data can leave the network. Ours cannot. A self-hosted agent that wires internal systems and cites sources is what compliance required.
Yes. It is the complete FIM One codebase, self-hosted. No usage limits, no feature gates, no expiry.
Enterprise adds: compliance tooling, dedicated support, SLA guarantees, and architecture review sessions. The core feature set is identical.
No. FIM One is source available: its license is not an OSI-approved open source license. The full codebase is on GitHub. You can read, fork, and self-host it. License terms are in the repository.
Yes. The license restricts deployment shape, not whether you charge money. You can deliver a FIM One-based solution to a single client as a dedicated, single-tenant deployment for their internal use, with FIM One branding kept intact. A separate commercial license is needed for white-labeling (shipping under the client's brand), serving multiple clients from one shared/multi-tenant deployment, or building a competing agent platform.
FIM One is provider-agnostic. It supports any OpenAI-compatible endpoint: GPT-4, Claude, Gemini, Mistral, Llama.