Project ARBR
Project ARBR is a Gyde-led open-source initiative that provides a self-hosted, provider-neutral AI gateway and control plane. It lets teams route, govern, observe, evaluate, and deploy model requests through one OpenAI-compatible endpoint, with human-approved, reversible routing decisions and measured cost/quality outcomes.
At a Glance
- Teams running AI traffic in production
- Developers building applications, agents, and chat interfaces on multiple LLM providers
- Enterprise AI/platform and operations teams needing spend governance and auditability
- Organizations seeking self-hosted, provider-neutral infrastructure
- +1 more
AI Tools by Project ARBR
(1)
ARBR
AI Gateway for LLM Teams
Discussions
No discussions yet
Be the first to start a discussion about Project ARBR
Latest News
Control-plane repository continued active development with click-to-sort dashboard table columns.
Python arbr-client 0.6.1 released on PyPI.
Arbr Control Plane 0.3.0 released with eval-backed routing, shadow/canary deployment, embeddings, realtime proxying, governance, security, and evidence reports.
Python arbr-client 0.5.0 released with the SDK release series accompanying the Control Plane 0.3 feature set.
Products & Services
Self-hosted AI gateway and control plane for multi-provider routing, spend tracking, budget governance, request observability, evaluation, recommendations, human approval, canary/shadow deployment, and rollback. It can operate standalone or in front of LiteLLM.
Zero-dependency Node client for chat, streaming, embeddings, model/provider discovery, status, retries, typed errors, attribution, and LangChain integration.
Python 3.11+ client with synchronous and asynchronous chat/streaming APIs, embeddings, provider/model discovery, usage analytics via read tokens, and optional LangChain support. The public PyPI page lists version 0.6.1 released August 17, 2026.
Standalone npm CLI that audits request logs for premium-model overuse or wraps a coding-agent session to measure cost without requiring a server, database, or signup.
Market Position
ARBR positions itself as an open-source, self-hosted control plane focused on the evidence-backed optimization loop—observe real traffic, discover savings, evaluate candidates, obtain human approval, roll out reversibly, and verify realized outcomes—rather than only being a multi-provider proxy. Its documented comparison set includes LiteLLM and managed AI gateways: ARBR can sit above LiteLLM or operate standalone, adding recommendations, evaluation gates, governance, approvals, and measured optimization.
Leadership
Founders
Prasanna Vaidya
Co-creator of Project ARBR and CEO of Gyde; his public GitHub profile describes his focus as AI systems research.
Shubham Deshmukh
Co-creator of Project ARBR, responsible for its architecture and engineering; publicly identified as Co-Founder & CTO at Gyde.
Vaibhav Domkundwar
Member of the three-person team identified in the launch materials as being behind ARBR; also listed on Product Hunt as part of the ARBR launch team.
Executive Team
Prasanna Vaidya
Co-creator; Gyde CEO
CEO at Gyde and an AI systems researcher; identified in ARBR launch material as part of the team behind the project.
Shubham Deshmukh
Co-creator; architecture and engineering lead
Public LinkedIn search results identify him as Co-Founder & CTO at Gyde and say he co-created ARBR and leads its architecture and engineering.
Founding Story
ARBR was built to address the production problem that LLM logs show cost but do not answer: which workloads can safely move to another model, what evidence supports the change, and whether the result held after rollout. Its initial vision was an open-source, self-hosted control layer where teams could observe real workloads, evaluate alternatives on representative traffic, approve changes, and verify savings without opaque autonomous switching.
Business Model
Revenue Model
The software is distributed as self-hosted open source under the MIT License. The public materials describe a free Product Hunt offering and no paid SaaS plans; Gyde separately provides consulting and implementation services related to routing, while Project ARBR itself is the open-source initiative.
Pricing Tiers
MIT-licensed Control Plane and published SDK/CLI packages; users run the gateway themselves. Product Hunt lists ARBR as Free.
Target Markets
- Teams running AI traffic in production
- Developers building applications, agents, and chat interfaces on multiple LLM providers
- Enterprise AI/platform and operations teams needing spend governance and auditability
- Organizations seeking self-hosted, provider-neutral infrastructure
- Teams using LiteLLM or other OpenAI-compatible gateways
- Routing production LLM workloads to an appropriate model by task, difficulty, quality, latency, policy, or cost
- Finding and validating safe model downgrades to reduce AI spend
- Operating a governed multi-provider AI gateway for applications, agents, and chat UIs
- Evaluating candidate models on representative internal traffic before rollout
- Shadow-testing and canary-deploying model changes with automatic rollback
- Auditing coding-agent or request-log traffic for premium-model overuse