liter-llm

Name: liter-llm
Author: Xberg.io

One MCP server for chat, embeddings, images, speech, OCR and batch jobs across many LLM providers

By Xberg.io Rust Local · stdio AI and ML

GitHub npm Website

48 popularity score ★ 264 ↓ 958/wk

Connect your client

Local (npm)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs OPENAI_API_KEY. Replace each <NAME> with your value.

Claude CodeTerminal

claude mcp add liter-llm -e OPENAI_API_KEY=<OPENAI_API_KEY> -- npx -y @xberg-io/liter-llm-cli mcp --config '<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>'

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "liter-llm": {
      "command": "npx",
      "args": [
        "-y",
        "@xberg-io/liter-llm-cli",
        "mcp",
        "--config",
        "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"
      ],
      "env": {
        "OPENAI_API_KEY": "<OPENAI_API_KEY>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "liter-llm": {
      "command": "npx",
      "args": [
        "-y",
        "@xberg-io/liter-llm-cli",
        "mcp",
        "--config",
        "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"
      ],
      "env": {
        "OPENAI_API_KEY": "<OPENAI_API_KEY>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "liter-llm": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@xberg-io/liter-llm-cli",
        "mcp",
        "--config",
        "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"
      ],
      "env": {
        "OPENAI_API_KEY": "<OPENAI_API_KEY>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.liter-llm]
command = "npx"
args = ["-y", "@xberg-io/liter-llm-cli", "mcp", "--config", "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"]

[mcp_servers.liter-llm.env]
OPENAI_API_KEY = "<OPENAI_API_KEY>"

Gemini CLITerminal

gemini mcp add --scope user -e OPENAI_API_KEY=<OPENAI_API_KEY> liter-llm npx -- -y @xberg-io/liter-llm-cli mcp --config '<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>'

Local (from source)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs OPENAI_API_KEY. Replace each <NAME> with your value.

FirstTerminal

cargo install --git https://github.com/xberg-io/liter-llm --rev f14fb8862f9cc73e2c4ab38a2876efe77fab8bf4 --locked liter-llm-cli

Claude CodeTerminal

claude mcp add liter-llm -e OPENAI_API_KEY=<OPENAI_API_KEY> -- liter-llm mcp --config '<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>'

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "liter-llm": {
      "command": "liter-llm",
      "args": [
        "mcp",
        "--config",
        "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"
      ],
      "env": {
        "OPENAI_API_KEY": "<OPENAI_API_KEY>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "liter-llm": {
      "command": "liter-llm",
      "args": [
        "mcp",
        "--config",
        "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"
      ],
      "env": {
        "OPENAI_API_KEY": "<OPENAI_API_KEY>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "liter-llm": {
      "type": "stdio",
      "command": "liter-llm",
      "args": [
        "mcp",
        "--config",
        "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"
      ],
      "env": {
        "OPENAI_API_KEY": "<OPENAI_API_KEY>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.liter-llm]
command = "liter-llm"
args = ["mcp", "--config", "<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>"]

[mcp_servers.liter-llm.env]
OPENAI_API_KEY = "<OPENAI_API_KEY>"

Gemini CLITerminal

gemini mcp add --scope user -e OPENAI_API_KEY=<OPENAI_API_KEY> liter-llm liter-llm -- mcp --config '<ABSOLUTE_PATH_TO_LITER_LLM_PROXY_TOML>'

Settings 3

`OPENAI_API_KEY`	Required: Key for the provider you call; set the matching one for others, such as ANTHROPIC_API_KEY.
`ANTHROPIC_API_KEY`	Optional: Key for Anthropic models, when you route requests there.
`LITER_LLM_MASTER_KEY`	Optional: Master key presented as a Bearer token in HTTP mode.

What it does

liter-llm is a Rust client library for calling many hosted and local language-model providers through one interface. Its command-line binary also carries an MCP server, started with `liter-llm mcp`, which turns that client into 22 tools an assistant can call: chat completions, embeddings, model listing, image generation, text-to-speech, transcription, moderation, reranking, provider-side search, OCR, provider file uploads, batch jobs and the Responses API.

The server reads the same `liter-llm-proxy.toml` as the project's OpenAI-compatible proxy, so models, aliases, caching, fallbacks and virtual keys defined there apply to tool calls too. It runs over stdio for a local client, or over Streamable HTTP on `POST /mcp` (port 3001 and address 127.0.0.1 by default) to share one server among several users. Besides tools it offers three prompts (summarize, translate, extract) and read-only resources for the configured models, the provider registry and per-model pricing.

In the v2.2.3 code, the file, batch and response tools are refused for anything but the master key context: a virtual key may only call the model-routed tools, and only for the models it lists (crates/liter-llm-proxy/src/mcp/mod.rs). Over stdio the whole server runs with the context set in the `[mcp]` table of the config.

What it needs

Runtime	Node.jsorRust

The liter-llm binary, version 2.2.3 from npm (`npx`, needs Node.js), Homebrew, Scoop, crates.io or a GitHub release; a `liter-llm-proxy.toml` whose `[mcp]` table sets `stdio_trust_local = true` or a `stdio_key_id`, which 2.2.3 needs to start over stdio; and an API key for each provider you call (an OpenAI, Anthropic or other account, billed by that provider). HTTP mode also needs a master or virtual key set in the config.

Limits

Released 2.2.3 exits on stdio start when the config sets neither `mcp.stdio_trust_local = true` nor `mcp.stdio_key_id`, so the docs' bare `liter-llm mcp` Claude Desktop example fails on that version; the trust-by-default fix exists only in a source build at a pinned commit. A `stdio_key_id` must name an existing virtual key or the server will not start. Image and audio tools return empty content when the chosen provider lacks the feature. Every tool returns its result as one JSON text block, with base64 for audio and file content.

Who it suits

Suits developers who already use several model providers and want an assistant to reach them through one configured gateway with aliases, caching and per-key model limits, or a team sharing one HTTP endpoint behind keys. It writes to provider accounts (uploads, deletes files, starts and cancels batches) and spends provider credit, so it fits people who manage those accounts. Someone who only needs one provider may prefer that provider's own tools; anyone on the released CLI must keep a config file in place.

Tools 22

chat: Send messages to a configured model and get a completion back
embed: Turn input text into embedding vectors with a chosen model
list_models: List the models available from the configured providers
generate_image: Create images from a text prompt
speech: Turn text into speech audio, returned as base64
transcribe: Convert base64 audio into text
moderate: Run text through a provider's moderation check
rerank: Order a list of documents by relevance to a query
search: Run a web or document search through a provider model
ocr: Pull text out of an image or document given by URL or base64
create_file: Upload a base64 file to the provider (master key only)
list_files: List files uploaded to the provider
retrieve_file: Get metadata for one uploaded file
delete_file: Remove an uploaded file from the provider
file_content: Download the raw content of an uploaded file
create_batch: Start a batch job from an uploaded input file
list_batches: List batch jobs
retrieve_batch: Get the status of one batch job
cancel_batch: Stop a batch job that is still running
create_response: Create a response through the Responses API
retrieve_response: Fetch a stored response by ID
cancel_response: Stop a response that is still in progress

Popularity

48 popularity score, on our list (the bar is 36)

★ 264 ↓ 958/wk

GitHub stars, whole repository	264	23.9 points
npm downloads, last week	958	23.9 points

It lives in the folder cli-proxy/npm of xberg-io/liter-llm, which holds more than this server. Its stars are the whole repository's, so in the score they never count for more than its own downloads.

Measured 2026-11-07. How the score works.

Where it is published

Code	xberg-io/liter-llm/cli-proxy/npm ★ 264, the whole repository
npm	@xberg-io/liter-llm-cli 2.7.0, 2026-11-06, the one the command runs
Container	ghcr.io/xberg-io/liter-llm

What it can reach

Runs	On your machine
Reaches	The internet, Your account at a service
Changes things	Yes: it can write, send, create or delete
Sign-in	An API key or token
Licence	MIT (npm)
Last code change	2026-11-06 (a release)

Runs on your machine and sends prompts, audio, images and files to whichever model providers the config names, using your provider API keys. Tools can upload and delete files on a provider account and start or cancel batch jobs and responses, all of which can cost money. Over stdio the server acts with the context the `[mcp]` table sets: with `stdio_trust_local = true` that is the master context, which unlocks every tool. HTTP mode binds 127.0.0.1 unless --host is given and answers 401 without a Bearer master or virtual key; its docs say not to expose it without TLS.

Over time

GitHub stars, whole repository 264 +2 (+1 %)

264261

13 Oct 20267 Nov 2026

npm downloads a week 958 +386 (+67 %)

2.6k48

6 Jul 20267 Nov 2026

Before the site measured it, read from npm's daily download counts.

Popularity score 48 −4 (−8 %)

5238

13 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-10-15.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![liter-llm on EveryMCPServer](https://everymcpserver.com/badge/liter-llm.svg)](https://everymcpserver.com/servers/liter-llm)

Connect your client

Add in one click

Or add it by command or config

Add in one click

Or add it by command or config

Settings 3

What it does

What it needs

Limits

Who it suits

Tools 22

Similar servers

llmwiki

LLM Checker

DeepSeek MCP Server

llmverify

Claudish

Ruflo