LiteLLM (YC W23)

Join our Town Hall on Tuesday, May 19th from 7:30-8:30am PST for security updates, product updates, and roadmap. Register here: forms.gle/GSegwaNRM3xBdh…

We're open sourcing LiteLLM Agent Platform Run Claude Code, Codex, Hermes or any coding agent in isolated K8s sandboxes The agent never sees your real API keys. Here's how 👇

5️⃣ new things on LiteLLM Agent Platform 📋 Skills Page - manage + attach agent skills ⚡ API Snippets - curl/python/js drawer in session view 🔧 EKS Watchdog - stuck sessions recover via pod DNS 🧹 Sandbox - dead pods killed instantly 🏥 Port Fix - fix node port resolution

5️⃣ new things on LiteLLM Agent Platform 🎉 Skills Library - reusable prompts for agents 🚀 EKS Deployments - web + worker in-cluster 💪 Sandboxes - faster replenish + health checks 🔥 Sandboxes - active sessions preempt warm pods 🧹 Agents UI - sortable table + paginated API

Join us next Monday, 5/18, at 7:30-8:30am PST for our May Town Hall. For agenda and registration, please visit: docs.litellm.ai/blog/may-townh…

Today we are launching the @LiteLLM Agent Platform - a simple, self-hosted infrastructure platform for running multiple agents in production. This is great for AI platform teams looking to enable their developers to ship agents without giving up control of their infra.

GPT-5.5 is now available on LiteLLM AI Gateway. Improved token efficiency. Stronger conceptual clarity. Excels at long-horizon deliverables. github.com/BerriAI/litell…

If you find other vulnerabilities, please report it here - github.com/BerriAI/litell… or email security@berri.ai

CVE-2026-30623 (vulnerability via Anthropic's MCP SDK) has been fixed since v1.83.6-nightly. Please refer to our blog post for more details. docs.litellm.ai/blog/mcp-stdio…

🔗 Blog post: docs.litellm.ai/blog/mcp-stdio…

Opus 4.7 is live on LiteLLM AI Gateway. More efficient tokens. Higher benchmark scores. Designed for complex, long-horizon work.

5️⃣ new things @LiteLLM 💪 vertex_ai - normalize finish_reason enum to OpenAI spec ✅ dashscope - preserve cache_control for prompt caching 🎉 TogetherAI - add together_ai/gpt-oss-120b 🛠️ ModelInfo - expose reasoning effort fields 🚀 PromptGuard - new guardrail integration

We are hosting our April Town Hall this Friday, April 10th at 7:30am PST. How to contribute: Add your thoughts to this ticket to help us shape the agenda - github.com/BerriAI/litell…

@SandySwipe @SethDeGroot Thanks @SandySwipe! We'll be hosting a public Town Hall on Friday, 4/10 as well: docs.litellm.ai/blog/april-tow…

@SethDeGroot @LiteLLM It's impressive to see how the @LiteLLM team handled that crisis with such expertise! 🔥 Incident response is crucial in today's tech landscape. A strong CI/CD pipeline ensures resilience. I'm excited to see how they continue to grow from this! 🤝

We’re investors in @LiteLLM. Yesterday they were hit by a highly sophisticated, multi-vendor supply chain attack. A 10-person team: • Contained it within hours • Engaged Mandiant for forensics • Rotated all credentials same-day • Migrated to OIDC Trusted Publishing Docker/proxy users were never affected. Startups are forged in the fire. You don't build a battle-hardened CI/CD pipeline without walking through the fire. Watching the 10-person team handle the crisis was a masterclass in incident response. This was their CloudFlare moment. Radical transparency + relentless execution. Proud to be on your cap table. 🤝

@kshitizh @mercor_ai Sharing links to our Security Update: docs.litellm.ai/blog/security-… and Town Hall: docs.litellm.ai/blog/security-…

@mercor_ai , the company training AI models for OpenAI, Anthropic, and Meta was hit via @LiteLLM, a widely used library for connecting apps to AI services. This is the part nobody talks about: Your security is only as strong as your weakest vendor.

A $10B AI startup just got breached. Not through phishing. Not through weak passwords. But, through an open-source library in their supply chain.

@building_KR @krrish_dh Hi @building_KR, appreciate the support! We are looking at April 11th

@krrish_dh @LiteLLM I’m a huge fan of LiteLLM. I have a quick question. The last stable version was released on March 16th (v1.82.3). When are you planning to release the next stable version?

5️⃣ new things @LiteLLM 🐳 Docker - Signed GHCR releases ✅ MCP - fix tool name prefix check 🛠️ S3 - use prepped URL for sigv4 signed releases 🚀 Presidio - fix anonymize text alignment on input ⚡️ Prometheus - add 7m + 10m latency buckets for long-running LLM Requests

📣 Announcing @LiteLLM's April Townhall Agenda: - Product roadmap - Reliability and security updates - Open Q&A with the team Register here: forms.gle/zyyHLhHEyWGo1A…

5️⃣ new things @LiteLLM 🐳 Docker - All Pip+NPM+Docker deps now pinned 💪 Qohash - new Guardrail provider 🛠️ Teams - enforce TPM/RPM limits for members 🎉 Gemini - skip fetch for File API URLs 🧹 Azure - forward api_version on embeddings

@shahin43 @garrytan @mercor_ai Sharing the updates from our public Town Hall: docs.litellm.ai/blog/security-…

@garrytan @mercor_ai this is so concerning about safety and security of Data. Pypi package leak, which poisoned @LiteLLM is main reason for this leak

Wow. Incredible amount of SOTA training data now just available to China thanks to @mercor_ai leak. Every major lab. Billions and billions of value and a major national security issue.