Charm
Also known as: Charmbracelet, Charm (charmbracelet), Crush
Charm makes the command line glamorous; its flagship Crush is a terminal native AI coding agent with best in class multi provider model flexibility and MCP extensibility.
Charm, the company behind the charmbracelet open source ecosystem, makes the command line glamorous with a suite of Go based terminal libraries and tools. Its flagship agentic product is Crush, a terminal native AI coding agent that connects to large language models and gives them tools to read, write, and execute code directly in the terminal. Crush stands out for serious model provider flexibility, letting developers choose among Anthropic, OpenAI, Gemini, Bedrock, Copilot, Vertex, Groq, and local models and switch mid session, with a Catwalk provider database that keeps model metadata current. It pulls in code intelligence through Language Server Protocols, reads project context files, persists sessions and history in SQLite, extends through Model Context Protocol servers and agent skills, and gives teams fine grained control over permissions and repo local configuration. Because it runs locally in the terminal it supports air gapped and local model workflows, and it works across macOS, Linux, Windows, and the BSDs.
Vendor details
Canonical URL
https://charm.land
Category
Coding agent
Funding status
Seed stage per the Crunchbase export, based in New York, founded by Christian Rocha. Specific round amount not retrieved this session. Charm is the company behind a widely adopted open source terminal tooling ecosystem, with Crush at over 150,000 GitHub stars.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Crush connects to LLM providers (Anthropic, OpenAI, Gemini, Bedrock, Copilot, Vertex, Groq, and local models), uses Language Server Protocols for code intelligence, extends via Model Context Protocol servers over stdio, http, and sse, and can call local tools like git, docker, and npm.
Sources & related URLs
Capability coverage
10.0 / 14 capabilities · 71%
| Integrations & Tool CallingCrush connects to many LLM providers, integrates Language Server Protocols for code intelligence, connects to MCP servers over stdio, http, and sse, and can directly call local tools such as git, docker, and npm per the Crush GitHub repo and blog. | Full |
|---|---|
| Workflow OrchestrationCrush is an agentic coding agent that reads, writes, and executes code, running multi step coding tasks with multiple work sessions and contexts per project per the Crush documentation. | Full |
| Knowledge Grounding & RAGCrush grounds its work in the codebase using Language Server Protocols for additional context and reads project context files (AGENTS.md, CRUSH.md, CLAUDE.md) plus session history to understand code, style, and conventions per the Crush AGENTS.md. | Full |
| Human Oversight & GuardrailsCrush gives teams fine grained control over permissions and repo local configuration, letting developers gate what the agent is allowed to do per the Crush review and documentation. | Full |
| Security, Identity & GovernanceCrush runs locally with permission controls, inspectable source available code, and file based secret handling, giving reasonable governance, but no named infosec certification applies to this local CLI tool. | Partial |
| Observability & AuditabilityCrush offers debug and debug_lsp options, tracks per model cost, and persists session history in SQLite that can be inspected, a developer level observability rather than a full observability suite. | Partial |
| Memory & State PersistenceCrush persists sessions and history in SQLite and maintains multiple work sessions and contexts per project per the Crush AGENTS.md. | Full |
| Deployment & Data ResidencyCrush runs locally in the terminal so code stays on the developer machine, supports air gapped environments by disabling provider auto update, works with local models, and runs on macOS, Linux, Windows, and the BSDs per the Crush repo. | Full |
| Prebuilt Agents / Templates / PacksCrush supports agent skills and reusable repo local configuration and context templates, but not a broad catalog of prebuilt agents. | Partial |
| Triggers & Channel CoverageCrush is an interactive terminal application invoked by the developer; no event triggers or multichannel coverage were documented this session. | Unable to verify |
| Model Flexibility & RoutingCrush offers best in class model provider flexibility across Anthropic, OpenAI, Gemini, Bedrock, Copilot, Vertex, Groq, and local models, lets developers switch models mid session, and auto updates the provider list from the open Catwalk database per the Crush repo. | Full |
| APIs / SDKs / MCP ExtensibilityCrush is extensible via Model Context Protocol servers over stdio, http, and sse, plus agent skills, and Charm ships open source Go libraries such as Fantasy, with repo local configuration and source available code per the Crush repo. | Full |
| Testing, Debugging & OptimizationCrush can execute code and offers debug tooling so developers can run and debug what the agent produces, but no formal built in agent evaluation framework was documented this session. | Partial |
| Browser / Computer-useCrush operates in the terminal and shell rather than driving a browser or graphical interface, so first class browser or GUI computer use was not documented this session. | Unable to verify |
Pricing
Free (source available); bring your own model keys
Cost watchouts
The tool is free, but running Crush means paying your chosen LLM provider for token usage, which scales with how much coding work you drive through it.
Variable cost rationale
Because developers bring their own model keys, the effective cost is LLM token usage paid to the chosen provider, which scales directly with coding activity.
Sales call required
No — self-serve available
Key ambiguities
Charm has not published a commercial or enterprise price for Crush; only the free, source available tool with bring your own model keys is documented.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- Aider — Open-source, model-agnostic terminal coding agent that works…
- Anthropic Claude Code — Anthropic's terminal and IDE-native coding agent with sub-agent…
- Augment Code — AI coding assistant evolving into an AI-native engineering platform
- Baz — Agentic coding platform whose AI agents review, govern, and secure…
- Blackbox AI — AI coding assistant for code generation and developer workflows