ShipCode

Plain-language coding from your terminal, powered by any AI model you want.

Why ShipCode

Model-agnostic. Point it at Ollama for free local inference, Claude for best-in-class reasoning, GPT for broad capability, Groq for speed, Gemini for massive context, or OpenRouter for everything at once. ShipCode doesn't care which model you use — it just ships code.

Limitless. Configure a fallback chain and ShipCode never hits a wall. Claude rate-limited? Jump to Groq. Groq down? Drop to local Ollama. Your flow never stops because one provider had a bad day.

Real file editing. ShipCode isn't a chat toy. The AI proposes structured shipcode-write and shipcode-edit blocks, shows you a diff, and waits for your approval before touching disk. You stay in control of every keystroke that lands in your repo.

Install

npm install -g shipcode-cli

Quick Start

shipcode setup    # configure your AI provider
shipcode          # open interactive shell

How It Works

  ╔═══════════════════════════════════════════╗
  ║   ShipCode  ·  v0.5.0                     ║
  ║   provider: anthropic → groq → ollama     ║
  ║   model: claude-sonnet-4.6                ║
  ╚═══════════════════════════════════════════╝

you > create src/utils/format-date.ts with a formatDate(date) helper

ai  > Sure — here's a small helper that handles ISO strings and
      Date objects, with a sensible default format.

      ```shipcode-write path="src/utils/format-date.ts"
      export function formatDate(input: string | Date): string {
        const d = typeof input === 'string' ? new Date(input) : input;
        return d.toISOString().slice(0, 10);
      }
      ```

  + src/utils/format-date.ts  (new file, 4 lines)

  Apply this change? [y/n] y
  ✓ Wrote src/utils/format-date.ts

you > _

Commands

| Command | Description | |---|---| | shipcode | Interactive AI shell | | shipcode setup | Configure providers and API keys | | shipcode chat "question" | One-shot AI query | | shipcode chat "question" -f file.ts | Query with file context | | shipcode doctor | Environment and provider health checks | | shipcode status | Project and provider status | | shipcode migrate | Import sources (directory, notion, obsidian, github) | | shipcode init | Initialize a project directory | | shipcode benchmark | Run performance benchmarks | | shipcode whoami | Show current account and tier |

Inside the interactive shell:

| Slash command | Description | |---|---| | /context file.ts | Add a file to the conversation | | /provider anthropic | Switch providers mid-session | | /chain | View or edit the fallback chain | | /save name | Save the current session | | /load name | Restore a saved session | | /exit | Leave the shell |

The Limitless Fallback Chain

Every provider has a bad day. Rate limits kick in at the worst moment, the API throws a 500, or you're on a plane with no WiFi. ShipCode's fallback chain means you don't care.

Configure a chain and ShipCode walks it top to bottom on every request. If one link fails, it silently moves to the next. Put a local Ollama model at the bottom and you have a free, offline safety net forever.

# ~/.shipcode/config.toml
[chain]
order = ["anthropic", "groq", "ollama"]

[providers.anthropic]
model = "claude-sonnet-4.6"

[providers.groq]
model = "llama-3.3-70b-versatile"

[providers.ollama]
model = "llama3.1:8b"

File Operations

When ShipCode wants to touch your files, it doesn't just do it. It proposes.

shipcode-write — create or overwrite a file:

```shipcode-write path="src/hello.ts"
export const hello = () => 'world';
```

shipcode-edit — patch an existing file with a diff:

```shipcode-edit path="src/hello.ts"
- export const hello = () => 'world';
+ export const hello = (name: string) => `hello, ${name}`;
```

ShipCode renders the diff, asks [y/n], and only writes if you approve. Say no and the suggestion evaporates.

Providers

| Provider | Models | Use Case | Cost | |---|---|---|---| | Ollama | llama3.1, gemma4, qwen2.5-coder, etc. | Privacy-first, offline, unlimited | Free (local) | | Anthropic | Claude Sonnet 4.6, Opus 4.7 | Best-in-class reasoning and code | Cloud API | | OpenAI | GPT-4o, GPT-4.1, o3-mini | Broad capability, reliable | Cloud API | | Groq | Llama 3.3 70B, Mixtral 8x7B | Ultra-fast inference | Cloud API | | Gemini | 2.0 Flash, 2.5 Pro | Massive context windows | Cloud API | | OpenRouter | Hundreds of models | One key, every model | Cloud API |

Mix and match. Stack them in a chain. ShipCode doesn't care.

Part of MakeShipHappen

ShipCode is the terminal layer of the MakeShipHappen suite:

• ShipSpace — AI-native development environment
• ShipMind — second brain for notes, code, and context
• ShipTalk — voice-to-text, everywhere on your Mac
• ShipCode — this tool

Four tools, one account, one subscription.

Pricing

• Free — 5 AI messages per month, every provider you bring (unlimited Ollama).
• Pro — $29/mo, unlimited AI messages, every feature.
• Team — $79/mo, 10 seats, shared billing.

See full pricing →

License

MIT