StartupUnited States
BerriAI (LiteLLM)
The gateway between your harness and every model provider.
BerriAI builds LiteLLM — the open-source proxy that normalizes 100+ LLM providers behind the OpenAI API with routing, budgets and observability. A quiet standard inside production agent stacks.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
Cons
- Independent testOn March 24, 2026, LiteLLM 1.82.7 and 1.82.8 on PyPI shipped code that stole credentials, cloud keys and secrets, likely through the Trivy GitHub Action compromise.[1][2]
- User reportsUsers report memory growth under load that requires restarts, and a LiteLLM job posting cites OOM issues.[1][2]
- User reportsUsers report that the proxy's in-memory API key cache is not cleared on config reload, so stale keys keep working.[1]
- MetadataThe repo has more than 1,700 open GitHub issues.[1][2]
- Best for
- Teams that need a self-hosted multi-provider LLM gateway with spend tracking.
- Watch out for
- After the March 2026 supply-chain incident, pin versions and verify package hashes.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
People
Co-founder & CTO
Harness engineering
Co-founder & CEO
Harness engineering
Open roles (18)
- Senior Backend Engineer
San Francisco · Full-time · Posted Nov 20, 2026
Engineering - Staff Backend Engineer
San Francisco · Full-time · Posted Nov 20, 2026
Engineering - ML Engineer
San Francisco · Full-time · Posted Nov 11, 2026
Engineering - Head of CS and Support
San Francisco · Full-time · Posted Nov 10, 2026
Engineering - Staff Engineer, AI
San Francisco · Full-time · Posted Nov 9, 2026
Engineering
Their open-source work
Python · ≈64k stars · Other
BerriAI/litellm
Call any LLM with one interface — the model gateway.
Maintained by BerriAI
v1.108.14releasedActivity
- Last push
- · today
- Latest release
- v1.108.14
· yesterday - Last 90 days
- 20+ releases
v1.108.141.108.14 (LTS)
· yesterday[!TIP] The lts image tag always points to the latest LTS release. Pull ghcr.io/berriai/litellm:lts to stay on the newest LTS patch Verify this release Same command for litellm-database, litellm-non_root and the docker.io/litellm/ images What's new in 1.108.14 (LTS) 3 PRs. 1…
v1.111.0-rc.3Pre-release
· yesterdayVerify this release Same command for litellm-database, litellm-non_root and the docker.io/litellm/ images What's new in 1.111.0-rc.3 15 PRs. 11 fixed. No action needed to upgrade Changes Anthropic: Fixed temperature=1 returning 400 on Claude models that removed sampling params…
- · yesterday
Verify this release Same command for litellm-database, litellm-non_root and the docker.io/litellm/ images What's new in 1.106.12 6 PRs. 2 fixed, 2 changed. No action needed to upgrade Changes Caching: Fixed GPT-6.2 system prompts missing the prompt cache on /v1/chat/completions…
- · yesterday
Verify this release Same command for litellm-database, litellm-non_root and the docker.io/litellm/ images What's new in 1.110.2 11 PRs. 8 fixed. No action needed to upgrade Changes Anthropic: Fixed temperature=1 returning 400 on Claude models that removed sampling params on…
v1.108.131.108.13 (LTS)
· yesterday[!TIP] The lts image tag always points to the latest LTS release. Pull ghcr.io/berriai/litellm:lts to stay on the newest LTS patch Verify this release Same command for litellm-database, litellm-non_root and the docker.io/litellm/ images What's new in 1.108.13 (LTS) 7 PRs. 5…
Checked against GitHub on Nov 21, 2026. Release notes are shortened to plain text.
On X
No posts from them in our X feed yet.
Neighbours in the directory
New York, NY · Est. 2021
Serverless cloud built for AI and agentic compute.
Modal lets teams run inference, fine-tuning and arbitrary agent workloads as serverless Python without managing infrastructure — a common substrate for harness backends at scale.
Seattle, WA · Est. 2019
Durable execution for long-running agent workflows.
Temporal's durable execution engine is widely adopted to make agent pipelines reliable: state survives crashes, retries are automatic, and workflows run for minutes or months. A key piece of serious harness infrastructure.
$430M total · $300M Series D (Feb 2026) at $5B valuation
San Francisco, CA · Est. 2021
Durable functions and workflows for AI agents.
Inngest runs agent logic as durable steps — queueing, retries, concurrency control and human-in-the-loop waits — without operating your own workflow infrastructure.
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
Maintains · 1
- LiteLLMProject
Similar entries · 4
- claude-code-proxy (fuergaosi233)Project
- CLIProxyAPIProject
- HeliconeStartup
- LiteLLMProject