Skip to content

AI Coding Assistants

Give your coding assistant a working knowledge of liter-llm. The plugin teaches your agent how to call any of the 174 providers, stream responses, use tools, and generate embeddings — so it writes correct liter-llm code the first time instead of guessing at the API.

The liter-llm plugin ships a set of agent skills, installed from this repo’s self-hosted marketplace (xberg-io/liter-llm):

  • Chat and streaming — build requests, handle streamed responses, and switch models across providers.
  • Tool calling — define tools and handle the model’s tool calls.
  • Embeddings — generate and use embeddings for search and retrieval.
  • MCP server — registers the liter-llm MCP server with your client. Use the compatible CLI described below.

Once installed, your assistant applies these skills automatically when you ask it to work with liter-llm.

Published CLI 2.2.3 predates the local-stdio trust and deferred-provider-credential fixes. Until a CLI release includes them, use a source build for startup without an extra auth config:

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

Put the installed binary on the assistant’s PATH. If a plugin has already cached an older binary, remove that plugin’s bin/liter-llm (or bin/liter-llm.exe) before retrying so it selects the source build. The released 2.2.3 binary requires mcp.stdio_trust_local = true or a configured mcp.stdio_key_id; configured providers may also require credentials at startup. The source build defers missing provider credentials until a request uses that provider. HTTP authentication remains configured separately.

Expand the section for your coding agent below.

Claude Code
/plugin marketplace add xberg-io/liter-llm
/plugin install liter-llm@liter-llm
Codex CLI
/plugins add https://github.com/xberg-io/liter-llm

Then search for liter-llm and select Install Plugin.

Cursor

Settings → Plugins → Add from URL → https://github.com/xberg-io/liter-llm, then select liter-llm.

Gemini CLI

From a checkout of this repository:

Terminal window
gemini extensions install ./plugin --skip-settings

The manifest is inside plugin/. See the Gemini extension reference for local installation and updates.

Configure the provider you use, for example gemini extensions config liter-llm OPENAI_API_KEY. Gemini filters sensitive shell variables; the extension declares these settings for secure credential injection. Skipping settings during installation avoids prompts for unused providers.

Kimi installation and compatibility

From a checkout of this repository, install with the Python Kimi CLI version used for native plugin validation:

Terminal window
uvx --from kimi-cli==1.51.0 kimi plugin install ./plugin

Native plugin installation was validated with Python Kimi CLI 1.51.0. The replacement @moonshot-ai/kimi-code 2.1.1 does not provide a plugin CLI subcommand; its interactive plugin loading has not been verified. These results do not establish compatibility with newer Kimi releases.

Factory Droid
droid plugin marketplace add https://github.com/xberg-io/liter-llm
droid plugin install liter-llm@liter-llm
GitHub Copilot CLI
copilot plugin marketplace add https://github.com/xberg-io/liter-llm
copilot plugin install liter-llm@liter-llm
Pi

Install from a local checkout now:

Terminal window
pi install ./plugin

After the next plugin release, install the same package used by OpenCode:

Terminal window
pi install npm:@xberg-io/opencode-liter-llm

Pi loads the bundled skills and prompt templates. Use /liter-llm-chat and /liter-llm-stream to start a task. Install the CLI separately as shown in the plugin README. This package does not register MCP servers, hooks, or executable extensions in Pi.

opencode

Add the package to opencode.json:

{
"$schema": "https://opencode.ai/config.json",
"plugin": ["@xberg-io/opencode-liter-llm"]
}

The MCP-capable integrations above install the same bundled MCP server and register it for you — there is nothing to configure by hand. The plugin points at scripts/mcp-launch.sh, which resolves the liter-llm binary at runtime (pnpm dlx @xberg-io/liter-llm-cli, uvx --from liter-llm-cli, Homebrew, or a downloaded release).

To register the server directly in any other MCP client (without the plugin), add it to that client’s MCP config:

{
"mcpServers": {
"liter-llm": {
"command": "npx",
"args": ["-y", "@xberg-io/liter-llm-cli@latest", "mcp", "--transport", "stdio"]
}
}
}

uvx --from liter-llm-cli liter-llm mcp --transport stdio works as an alternative command.

For the released 2.2.3 CLI, the stdio transport requires the auth binding described above. The source build trusts local stdio by default. See the MCP server guide for transport and provider configuration.

For the Hermes runtime, install the plugin from PyPI:

Terminal window
pip install liter-llm-hermes-plugin

The plugin, its skills, and support for more agents are maintained in this repository (xberg-io/liter-llm).

The generated bundle targets Claude Code, Codex, Cursor, Gemini CLI, Kimi, OpenCode, Factory Droid, Hermes, Pi, Agent Plugins and GitHub Copilot CLI. Use the plugin installation guide for the runtime’s installation method and configuration.

Runtime Integration
Claude, Codex, Cursor, Factory Native plugin manifests, skills and MCP configuration
Gemini CLI Extension manifest, command templates and MCP configuration
Kimi Plugin manifest and session instructions
OpenCode npm adapter, instructions and MCP configuration
Hermes Python plugin, explicitly selected skills and a local stdio MCP bridge
Pi Skills and prompt templates; execute operations through the CLI
Agent Plugins, Copilot Agent Plugins 1.0 manifest, skills and MCP configuration

Install the product CLI before using Hermes or Pi. Hermes starts the installed CLI directly; its wheel does not include the auto-installing launcher used by other MCP clients. Select list_tools through the product’s Hermes bridge to inspect the supported tools and request schemas. Select skills explicitly when needed; they are not appended automatically to every conversation.

From a checkout, install the generated Hermes package into Hermes’ Python environment:

Terminal window
pip install ./plugin/.hermes/package

Enable the plugin in Hermes configuration. Loading project plugins also requires HERMES_ENABLE_PROJECT_PLUGINS=true. After a release publishes the updated package, the registry install is pip install liter-llm-hermes-plugin.

Pi resources can be installed from the checkout with pi install ./plugin. After a plugin release includes them, install npm:@xberg-io/opencode-liter-llm through Pi. Pi does not register an MCP server, hooks or executable extensions from this package.

The public skills directory contains instructions you can install independently of the runtime plugin. Use the skills CLI to list and select them:

Terminal window
npx skills add xberg-io/liter-llm --list
npx skills add xberg-io/liter-llm

Choose the target agents interactively. For a local checkout, use npx skills add ./skills. This installs instructions; it does not install the product CLI, enable MCP, download models or set provider credentials.

The local stdio MCP server trusts its harness by default and starts without provider keys. Configure provider keys before making provider-backed requests; a missing key produces a request error. HTTP MCP clients must still supply a master or virtual key. Use liter-llm api for the proxy and liter-llm mcp for the tool server. Chat and streaming are SDK, proxy or MCP operations, rather than additional CLI subcommands.