Tabby
Tabby is a self-hosted AI coding assistant · Copilot-shaped autocomplete you run on your own hardware for privacy.
Tabby is a self-hosted AI coding assistant · Copilot-shaped autocomplete you run on your own hardware for privacy.
Basic
Tabby launched from TabbyML · an open-source self-hostable coding assistant. Deploy with Docker on a server with a GPU; developers get IDE autocomplete via IDE plugins (VSCode, JetBrains, Vim). Target customers: enterprises with air-gapped environments where Copilot/Cursor are forbidden.
Deep
Tabby ships pre-configured with StarCoder2 and other open models. Admins pick the model based on GPU budget · a single A100 serves ~10-20 concurrent developers. Features beyond autocomplete include chat, code-search, and a team dashboard with analytics. Enterprise tier adds SSO, audit logs, and on-prem LDAP. Apache-2.0 with a paid enterprise offering.
Expert
Tabby uses Rust + Python for the inference server · wraps llama.cpp or vLLM depending on hardware. Context retrieval uses local repo indexing (tree-sitter + embeddings) so suggestions are repo-aware. The enterprise tier is the commercial moat · compliance, audit, multi-team. Competitors in self-hosted: Refact.ai, Continue (with Ollama), CodeGeeX. Tabby's Rust-native server gives it a performance edge.
Depending on why you're here
- ·Apache-2.0 self-hosted autocomplete server
- ·Rust + Python · wraps llama.cpp / vLLM
- ·Uses StarCoder2 and similar open code models
- ·Deploy with Docker + one A100 · serves 10-20 devs
- ·Target: air-gapped orgs or privacy-conscious teams
- ·Compare: Continue + Ollama (simpler) vs Tabby (managed, analytics)
- ·TabbyML raised seed/Series A · commercial strategy via enterprise tier
- ·Niche but sticky · enterprises that can't use Copilot
- ·Competes with Refact.ai, Continue, GitHub Copilot for Business
- ·A self-hosted alternative to Copilot
- ·Runs on your own server · no code leaves your network
- ·Popular with banks, defense, healthcare
Tabby is the answer to "we want Copilot but compliance said no." Growing steadily where privacy matters.