LLM Keys & Endpoints

Complete reference for LLM provider credentials used with LibreChat. All keys go in ~/LibreChat/.env (or via augur env).

At least one LLM provider is required. You can use multiple simultaneously.

Quick Setup

Native providers (OpenAI, Anthropic): just set the key in .env — auto-detected.

Sign up at the provider URL
Copy your API key
Add it to ~/LibreChat/.env (via augur env)
Restart: augur restart

Custom providers (Groq, Gemini, Mistral, etc.): need a YAML endpoint block too.

Set the key in .env
Add the endpoint block to librechat-user.yaml (via augur yaml)
Restart: augur restart

Free Tier Providers

All providers below offer free API tiers (rate-limited, no billing required). Add the endpoint block to librechat-user.yaml (augur yaml) to enable.

Reference: cheahjs/free-llm-api-resources

Groq -- Fastest free inference


Signup	https://console.groq.com/keys
Free limits	14,400 req/day (varies by model), 6K-30K tokens/min
Best models	`openai/gpt-oss-120b`, `llama-3.3-70b-versatile`, `llama-3.1-8b-instant`, `kimi-k2-0905`
Env var	`GROQ_API_KEY=gsk_...`
Notes	Custom LPU hardware, extremely fast. Best free option for daily use.

Google Gemini -- Largest free context


Signup	https://aistudio.google.com/apikey
Free limits	15 RPM, 1M tokens/day (Gemini Flash)
Best models	`gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-3-flash-preview`
Env var	`GEMINI_API_KEY=AI...`
Notes	Very generous token limits. 1M context on paid tier.

Mistral -- Best European provider


Signup	https://console.mistral.ai/api-keys
Free limits	1 req/sec, 500K tokens/min
Best models	`mistral-small-latest`, `mistral-large-latest`, `magistral-small-latest`, `magistral-medium-latest`
Env var	`MISTRAL_API_KEY=...`
Notes	Codestral is great for code tasks. Free tier very generous.

Cerebras -- Fast open-model inference


Signup	https://cloud.cerebras.ai
Free limits	30 RPM, 14,400 req/day
Best models	`qwen3-235b`, `gpt-oss-120b`, `llama3.1-8b`
Env var	`CEREBRAS_API_KEY=csk-...`
Notes	Very fast inference on open models.

Cohere -- Command R family


Signup	https://dashboard.cohere.com/api-keys
Free limits	20 RPM, 1,000 req/month
Best models	`command-a-03-2025`, `command-a-reasoning-08-2025`, `command-r7b-12-2024`
Env var	`COHERE_API_KEY=...`
Notes	Good for RAG use cases. Lower monthly limit but capable models.

GitHub Models -- Free via GitHub PAT


Signup	https://github.com/settings/tokens (PAT with no scopes needed)
Free limits	Rate-limited, generous for personal use
Best models	`gpt-4o-mini`, `o4-mini`, `Phi-4`
Env var	`GITHUB_MODELS_PAT=ghp_...`
Notes	Uses your GitHub account. Access via GitHub Marketplace Models.

Alibaba Cloud / Qwen -- Free 1M tokens/month


Signup	https://dashscope.console.aliyun.com/apiKey
Free limits	1M free tokens/month, rate-limited
Best models	`qwen3-max`, `qwen-plus`, `qwen-flash`, `qwen-long`
Env var	`DASHSCOPE_API_KEY=sk-...`
Notes	Alibaba's Qwen family. International endpoint. Landing page.

OpenRouter -- Aggregator with free models


Signup	https://openrouter.ai/keys
Free limits	20 RPM, 50 req/day (free models only, `:free` suffix)
Best models	`google/gemini-2.5-flash:free`, `deepseek/deepseek-r1:free`, `meta-llama/llama-3.3-70b-instruct:free`
Env var	`OPENROUTER_API_KEY=sk-or-...`
Notes	Aggregates many providers. Free models marked with `:free`. Also has paid models (see below).

Recommended Free Combo

For a free multi-model LibreChat setup, use these 3 together:

Groq -- daily driver (fast, high limits)
Gemini -- large context tasks (1M tokens/day)
Mistral -- reasoning tasks (Magistral)

Paid Providers

OpenAI -- GPT-4o, o1, o3


Signup	https://platform.openai.com/api-keys
Pricing	Pay-per-token, prepaid credits
Best models	`gpt-4o`, `o4-mini`, `gpt-4o-mini`
Env var	`OPENAI_API_KEY=sk-...`
Notes	Native LibreChat endpoint (no custom config needed). Set in `.env` and it works.

Anthropic -- Claude Opus, Sonnet, Haiku


Signup	https://console.anthropic.com/settings/keys
Pricing	Pay-per-token, prepaid credits
Best models	`claude-opus-4-6`, `claude-sonnet-4-6`, `claude-haiku-4-5-20251001`
Env var	`ANTHROPIC_API_KEY=sk-ant-...`
Notes	Native LibreChat endpoint (no custom config needed). Set in `.env` and it works.

OpenRouter -- Multi-provider gateway (paid tier)


Signup	https://openrouter.ai/keys
Pricing	Pay-per-token, $10+ credit gives 1000 req/day
Best models	All OpenAI, Anthropic, Google, Meta models via single key
Env var	`OPENROUTER_API_KEY=sk-or-...`
Notes	Recommended if you want access to all major models through one key. Same key works for free `:free` models too.

Google Gemini -- Paid tier


Signup	https://aistudio.google.com/apikey
Pricing	Pay-per-token (same key as free tier, billing enabled)
Best models	`gemini-2.5-pro`, `gemini-2.5-flash` (higher limits)
Env var	`GEMINI_API_KEY=AI...`
Notes	Same key as free tier. Enable billing for higher rate limits and pro models.

Mistral -- Paid tier


Signup	https://console.mistral.ai/api-keys
Pricing	Pay-per-token
Best models	`mistral-large-latest`, `magistral-medium-latest` (higher limits)
Env var	`MISTRAL_API_KEY=...`
Notes	Same key as free tier. Enable billing for higher limits and large model access.

Claude Max Subscription

Use your Claude Pro/Max subscription as a LibreChat endpoint -- no per-token billing.


Requires	Active Claude Pro or Max subscription
Setup	`augur proxy setup` (installs CLIProxyAPI, registers service)
How it works	CLIProxyAPI runs locally on `:8317`, translates OpenAI-compatible requests to Claude CLI
Env var	`CLAUDE_CODE_OAUTH_TOKEN=sk-ant-oat01-...` (in `~/.claude-auth.env`)
Full guide	docs/claude-token-wrapper.md

Add the "Claude Max" endpoint block to librechat-user.yaml (augur yaml) and restart.

GitHub Secrets

Store LLM keys as GitHub repository secrets to enable CI validation. Go to: Settings > Secrets and variables > Actions > New repository secret

Secret name	Used by
`OPENAI_API_KEY`	LLM key check CI job
`ANTHROPIC_API_KEY`	LLM key check CI job
`OPENROUTER_API_KEY`	LLM key check CI job
`GROQ_API_KEY`	LLM key check CI job
`GEMINI_API_KEY`	LLM key check CI job
`MISTRAL_API_KEY`	LLM key check CI job

All are optional -- the CI job skips providers whose secrets aren't set.

Example .env

# ── Paid (pick at least one) ──
# OPENAI_API_KEY=sk-...
# ANTHROPIC_API_KEY=sk-ant-...

# ── OpenRouter (one key, all models) ──
# OPENROUTER_API_KEY=sk-or-...

# ── Free providers (uncomment the ones you signed up for) ──
GROQ_API_KEY=gsk_abc123...
GEMINI_API_KEY=AIza...
# MISTRAL_API_KEY=...
# CEREBRAS_API_KEY=csk-...
# COHERE_API_KEY=...
# GITHUB_MODELS_PAT=ghp_...
# DASHSCOPE_API_KEY=sk-...

# Enable custom endpoints
ENDPOINTS=openAI,anthropic,custom

Then add the corresponding endpoint blocks to librechat-user.yaml (augur yaml) and restart.

Endpoint YAML Blocks

Custom providers need an endpoint block in librechat-user.yaml (augur yaml). Copy the block(s) you need below. Set the matching key in .env first.

Native providers (OpenAI, Anthropic) need NO YAML — just the .env key.

# Paste into librechat-user.yaml via: augur yaml
endpoints:
  custom:

    # ── Groq (free, very fast inference) ──
    - name: "Groq"
      apiKey: "${GROQ_API_KEY}"
      baseURL: "https://api.groq.com/openai/v1"
      models:
        default: [openai/gpt-oss-120b, llama-3.3-70b-versatile, llama-3.1-8b-instant, kimi-k2-0905]
        fetch: false  # true fetches all provider models — use false for curated list
      titleConvo: true
      titleModel: "llama-3.1-8b-instant"
      modelDisplayLabel: "Groq"

    # ── Google Gemini (free tier, 15 RPM / 1M TPD) ──
    - name: "Gemini"
      apiKey: "${GEMINI_API_KEY}"
      baseURL: "https://generativelanguage.googleapis.com/v1beta/openai/"
      models:
        default: [gemini-2.5-flash, gemini-2.5-pro, gemini-3-flash-preview]
        fetch: false
      titleConvo: true
      titleModel: "gemini-2.5-flash"
      modelDisplayLabel: "Gemini"

    # ── Mistral (free tier: mistral-small) ──
    - name: "Mistral"
      apiKey: "${MISTRAL_API_KEY}"
      baseURL: "https://api.mistral.ai/v1"
      models:
        default: [mistral-small-latest, mistral-large-latest, magistral-small-latest, magistral-medium-latest]
        fetch: false
      titleConvo: true
      titleModel: "mistral-small-latest"
      modelDisplayLabel: "Mistral"

    # ── Cerebras (free, 30 RPM / 14.4K RPD) ──
    - name: "Cerebras"
      apiKey: "${CEREBRAS_API_KEY}"
      baseURL: "https://api.cerebras.ai/v1"
      models:
        default: [qwen3-235b, gpt-oss-120b, llama3.1-8b]
        fetch: false
      titleConvo: true
      titleModel: "llama3.1-8b"
      modelDisplayLabel: "Cerebras"

    # ── Cohere (free, 20 RPM / 1K req/month) ──
    - name: "Cohere"
      apiKey: "${COHERE_API_KEY}"
      baseURL: "https://api.cohere.com/compatibility/v1"
      models:
        default: [command-a-03-2025, command-a-reasoning-08-2025, command-r7b-12-2024]
        fetch: false
      titleConvo: true
      titleModel: "command-r7b-12-2024"
      modelDisplayLabel: "Cohere"

    # ── GitHub Models (free, uses GitHub PAT) ──
    - name: "GitHub Models"
      apiKey: "${GITHUB_MODELS_PAT}"
      baseURL: "https://models.inference.ai.azure.com"
      models:
        default: [gpt-4o-mini, o4-mini, Phi-4]
        fetch: false
      titleConvo: true
      titleModel: "gpt-4o-mini"
      modelDisplayLabel: "GitHub Models"

    # ── Alibaba Cloud / Qwen (free 1M tokens/month) ──
    - name: "Qwen"
      apiKey: "${DASHSCOPE_API_KEY}"
      baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
      models:
        default: [qwen3-max, qwen-plus, qwen-flash, qwen-long]
        fetch: false
      titleConvo: true
      titleModel: "qwen-flash"
      modelDisplayLabel: "Qwen"

    # ── OpenRouter (some free models with :free suffix) ──
    - name: "OpenRouter"
      apiKey: "${OPENROUTER_API_KEY}"
      baseURL: "https://openrouter.ai/api/v1"
      models:
        default:
          - "google/gemini-2.5-flash:free"
          - "nvidia/nemotron-3-super-120b:free"
          - "meta-llama/llama-3.3-70b-instruct:free"
          - "deepseek/deepseek-r1:free"
        fetch: false
      titleConvo: true
      titleModel: "meta-llama/llama-3.3-70b-instruct:free"
      modelDisplayLabel: "OpenRouter"

    # ── Claude Max (subscription via CLIProxyAPI) ──
    # Requires: augur proxy setup + token
    # See docs/claude-token-wrapper.md for full guide.
    # - name: "Claude Max"
    #   apiKey: "dummy"
    #   baseURL: "http://localhost:8317/v1"
    #   models:
    #     default: [claude-sonnet-4-6, claude-opus-4-6, claude-haiku-4-5-20251001]
    #     fetch: false
    #   titleConvo: true
    #   titleModel: "claude-sonnet-4-6"
    #   directEndpoint: true
    #   summarize: false

Only include the providers you have keys for. Delete blocks you don't need.

Adding More Providers

Any OpenAI-compatible API can be added as a custom endpoint. See the LibreChat docs and these resources for more options:

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

LLM Keys & Endpoints

Quick Setup

Free Tier Providers

Groq -- Fastest free inference

Google Gemini -- Largest free context

Mistral -- Best European provider

Cerebras -- Fast open-model inference

Cohere -- Command R family

GitHub Models -- Free via GitHub PAT

Alibaba Cloud / Qwen -- Free 1M tokens/month

OpenRouter -- Aggregator with free models

Recommended Free Combo

Paid Providers

OpenAI -- GPT-4o, o1, o3

Anthropic -- Claude Opus, Sonnet, Haiku

OpenRouter -- Multi-provider gateway (paid tier)

Google Gemini -- Paid tier

Mistral -- Paid tier

Claude Max Subscription

GitHub Secrets

Example .env

Endpoint YAML Blocks

Adding More Providers

FilesExpand file tree

llm-keys.md

Latest commit

History

llm-keys.md

File metadata and controls

LLM Keys & Endpoints

Quick Setup

Free Tier Providers

Groq -- Fastest free inference

Google Gemini -- Largest free context

Mistral -- Best European provider

Cerebras -- Fast open-model inference

Cohere -- Command R family

GitHub Models -- Free via GitHub PAT

Alibaba Cloud / Qwen -- Free 1M tokens/month

OpenRouter -- Aggregator with free models

Recommended Free Combo

Paid Providers

OpenAI -- GPT-4o, o1, o3

Anthropic -- Claude Opus, Sonnet, Haiku

OpenRouter -- Multi-provider gateway (paid tier)

Google Gemini -- Paid tier

Mistral -- Paid tier

Claude Max Subscription

GitHub Secrets

Example .env

Endpoint YAML Blocks

Adding More Providers