Claude Code Alternatives in 2026
For local models, no subscription, or both.
Claude Code is one of the best coding agents around. It also needs a paid Claude plan and runs Claude models only. If you want local models or a tool that costs nothing, here are the options that do that today, checked against their own docs on October 30, 2026. Bodega One Code is one of them, and it can run Claude Code too.
Claude Code alternatives that run local models.
Every tool here can run a model on your own machine. Prices and terms change often, so check each link before you commit.
| Tool | License and price | Local models | Notes |
|---|---|---|---|
| Bodega One Code | Free to use, source not public | Yes: Ollama, LM Studio, llama.cpp, any OpenAI-compatible server | Desktop IDE with an agent, air-gap mode and pre-landing checks. Can also run Claude Code over ACP. |
| OpenCode | MIT, free | Yes: Ollama, LM Studio, llama.cpp | Terminal agent with 75+ providers. Optional paid model tiers. |
| Codex CLI | Apache-2.0 | Yes: --oss for Ollama or LM Studio, or custom providers | OpenAI's terminal agent. CLI access comes with ChatGPT Plus ($20) and up, or an API key. |
| Aider | Apache-2.0, free | Yes: almost any LLM, including Ollama | Terminal pair programmer built around git. Works with Ollama and other local models. |
| Cline | Apache-2.0, free | Yes: Ollama, LM Studio, any OpenAI-compatible API | Runs in VS Code, JetBrains and the terminal. |
| GitHub Copilot CLI | Proprietary, Free plan included | Yes: BYOK, including Ollama (the model must support tool calling and streaming) | Included on Copilot Free, Pro ($10) and up, metered in AI credits. |
Can Claude Code run local models?
It works, but Anthropic does not support it. Claude Code talks to whatever server ANTHROPIC_BASE_URL points at, as long as it speaks the Anthropic Messages API. Anthropic's own gateway docs add that it does not support routing Claude Code to non-Claude models through any gateway, so features can break between releases.
Two local servers speak that API today. Ollama added it in v0.14.0 and recommends a context window of 64K or more and a model with tool support. Strata, which runs large mixture-of-experts models on 12 to 24 GB GPUs, writes the whole Claude Code setup for you with one command. Our strata connect guide walks through it, including the global settings file it changes.
Mind the context. Strata measured Claude Code's full system prompt and 20 tools at about 15.6K tokens, so a local model with a small window has little room left for your code. Set CLAUDE_CODE_MAX_CONTEXT_TOKENS to the window your server really serves, or Claude Code assumes a larger one and compacts at the wrong time.
If you want local models as a supported, everyday choice, a tool built around BYOLLM is the simpler path. See the best local LLMs for coding by VRAM tier.
Claude Code vs Bodega One Code, side by side.
One is a subscription agent tied to one model family. The other is a free IDE that runs whatever model you bring.
Claude Code
- An agentic coding tool for the terminal, VS Code and JetBrains, a desktop app and the web. Strong defaults: Opus 5.5 on every paid plan.
- Needs a paid Claude plan (Pro from $17 a month billed annually, Max $100, Ultra $200, Team and Enterprise per seat) or pay-as-you-go API billing. Not on the Free plan.
- Claude models only. Anthropic supports its API, Amazon Bedrock, Google Cloud, Microsoft Foundry and LLM gateways, but not routing to non-Claude models.
- Prompts and model outputs go over the network to the model provider. Plan usage is capped per 5-hour and weekly window (Ultra drops the 5-hour cap); API use is billed per token.
- Source available on GitHub, but not open source: "All rights reserved" under Anthropic's Commercial Terms.
Bodega One Code
- A standalone desktop IDE: Monaco editor, AI chat and an autonomous agent in one app for Windows, macOS and Linux.
- BYOLLM: 10+ provider presets including Anthropic, plus local models through Ollama, LM Studio, llama.cpp or any OpenAI-compatible server.
- Local-first with an air-gap mode. With a local model, no code leaves your machine.
- QEL (Quality Enforcement Layer) checks the code the agent writes before the change lands.
- Free to use, commercial work included. No subscription and no usage window.
What Claude Code does well, and where each one fits.
If Claude models are what you want and the plan price works for you, Claude Code is a great pick. The case for an alternative is about models, cost and where your code goes.
Pick Claude Code if
You want Claude models with the least setup and a plan covers your usage.
- Claude models, tuned by the team that builds them. Opus 5.5 is the default on every paid plan.
- Runs everywhere you work: terminal, VS Code, JetBrains, desktop, web and mobile.
- A mature ecosystem: MCP servers, hooks, skills, sub-agents and managed settings for teams.
- Enterprise controls: SSO, gateways, managed policy and zero data retention for qualified organizations.
Pick Bodega One Code if
You want any model, local included, in a free app on your machine.
- You want to run local or open-weight models as a first-class choice, not through an unsupported gateway.
- You want no subscription and no usage window for the tool itself.
- You want an air-gap mode where no code leaves the machine.
- You like Claude Code and want to keep it: Bodega can run it as one agent among several over ACP.
You do not have to pick one
Bodega One Code runs external agents over ACP. Install Claude Code's ACP adapter, sign in to Claude Code once in your terminal, and pick Claude Code as the harness for a chat or a Fleet session. A Bodega turn on a local model can follow a Claude Code turn in the same thread. Claude Code still uses Anthropic's models and billing, its edits skip QEL but still go through Bodega's approvals, and ACP agents are off while air-gap mode is on.
Coming from
Claude Code.
Can Claude Code use local models?+
Not officially. Anthropic's gateway docs say it does not support routing Claude Code to non-Claude models through any gateway. In practice, any server that speaks the Anthropic Messages API can be set as ANTHROPIC_BASE_URL: Ollama (v0.14.0 and later) and Strata both do. Expect some features to break, and give the model at least a 32K context window, because Strata measured Claude Code's full prompt and tools at about 15.6K tokens.
Is there a free Claude Code alternative?+
Yes. OpenCode (MIT), Aider, Cline and Codex CLI (Apache-2.0) are free to run, and you pay only your model provider, or nothing with a local model. Bodega One Code is a free desktop IDE, commercial work included. Claude Code itself is not on the Claude Free plan; it starts with Pro at $17 a month billed annually or $20 monthly.
Is Claude Code open source?+
No. The license file in the anthropics/claude-code repository reads "All rights reserved" and points to Anthropic's Commercial Terms of Service. Codex CLI, Aider, Cline and Gemini CLI are Apache-2.0, and OpenCode is MIT.
Does Claude Code send my code to Anthropic?+
Claude Code runs on your machine but, in Anthropic's words, sends data over the network, including all user prompts and model outputs. On consumer plans, training on that data depends on a setting; commercial plans are not trained on by default and keep data for 30 days. With a local model and air-gap mode on, Bodega One Code blocks the agent's web tools, network commands, updates and cloud features.
Can I keep using Claude Code inside Bodega One Code?+
Yes. Bodega One Code runs external coding agents over ACP (Agent Client Protocol). Install the claude-code-acp adapter and sign in to Claude Code once in your terminal (claude auth login), and Claude Code can run inside a Bodega chat or as a Fleet session, with your project rules and memory going in and its edits going through Bodega's approvals. ACP agents do not go through Bodega's Quality Enforcement Layer, they are switched off while air-gap mode is on, and Claude Code still uses Anthropic's models and your Claude plan or API billing.
Any model. Your machine. No plan required.
The editor, the autonomous agent, air-gap mode and verification in one desktop app. Bring Claude, a local model, or both. Free to use, commercial work included.
Download FreeFree to use, commercial work included. In open beta now. Windows, macOS, Linux.