Guides
Choose a workflow, not a list of flags. Start with everyday tasks or the browser workspace, then explore agents, automation, and deployment as you need them.
Everyday workflows
Ask questions, understand code, and make useful changes.
Usage
Core command usage for `exec`, `ask`, `chat`, flags, examples, and agent selection.
File editing
Edit files with natural-language instructions, targeted ranges, and multiple diff formats.
Shell integration
Alias and shell-completion setup for using term-llm comfortably from the command line.
Search
Use web search in term-llm, choose external providers, and control native versus external search routing.
Browser workspace
A complete interface for conversations, projects, and agent work.
Agents and automation
Give repeatable work its own instructions, tools, and schedule.
Agents
Use built-in agents or create your own workflow bundles with their own provider, model, tools, and instructions.
Skills
Use and manage portable instruction bundles that add task-specific context.
Memory
Mine durable facts from completed sessions, search fragments, manage insights, and understand how memory differs from sessions.
Autonomous loops
Run an agent in a loop until a done condition is met.
Job runner
Run the jobs server, define scheduled work, and manage job definitions and runs from the CLI or API.
Notifications
Send notifications through Telegram or web push from the command line.
Models and tools
Connect external tools and build on the runtime.
MCP servers
Add external tools via Model Context Protocol and use them from term-llm commands.
Serving tools via MCP
Run term-llm as an MCP server over HTTP, exposing file, search, shell, and web tools to any MCP client.
Text embeddings
Generate vector embeddings for search, RAG, clustering, semantic similarity, and retrieval workflows.
Web and deployment
Run on your own infrastructure and reach your agents remotely.
Hub
Run term-llm Hub: one dashboard over many term-llm web nodes, with server-side token injection and one-click open.
Discourse Vibe (dv) + Hub
Connect dv (Discourse Vibe) dev containers to a term-llm Hub as reverse nodes: the reverse-node contract term-llm exposes, the dv-hooks reference implementation, and a troubleshooting guide for the common 'waiting for reverse connection' traps.
Agent Containers
Create isolated, persistent term-llm agents with the built-in Docker Compose workspace manager.
Telegram Bot
Run term-llm as a Telegram bot: create a bot, configure access control, and chat with your agent from any device.
WebRTC direct routing
Bypass a relay server and connect the browser directly to a home-hosted term-llm instance over a WebRTC data channel.
Media
Work with images, speech, music, and video.
Image generation
Generate and edit images with Gemini, OpenAI, ChatGPT, xAI, Venice, Flux, or OpenRouter.
Audio generation
Generate speech audio with Venice AI, Gemini, and ElevenLabs text-to-speech models.
Transcription
Transcribe audio files to text with OpenAI, Mistral Voxtral, Venice, ElevenLabs, a local Whisper server, or whisper.cpp CLI.
Music generation
Generate music and sound effects with Venice AI and ElevenLabs.
Video generation
Generate videos with Venice AI using text-to-video or image-to-video models.
Troubleshooting and internals
Diagnose issues and understand runtime behavior.
Debugging
Use provider debug output and debug logs to figure out what the runtime is actually doing.
Benchmarking providers
Measure client-observed LLM latency, decode throughput, and effective prefill performance.
Terminal-host lifecycle
Publish visible chat state safely to Herdr, cmux, tmux, Zellij, and other terminal hosts.