open source · MIT · npm

A coding agent
that lives in your
terminal.

Santra is a repository-aware coding assistant. It reads your codebase, reasons about it, and makes precise edits — using whichever model you choose.

npm install -g santra-cli GitHub →

santra / my-project↑↓ scroll · Enter send · / commands · F2 select · Esc stop

Welcome to Santra

Santra is a repository-aware coding assistant for the terminal.

docs → santra-cli.dev/docs

github → github.com/your-org/santra-cli

discord→ discord.gg/santra

Tips: press Enter to send your current prompt.

›Ask the agent anything… (/ for commands)

◆qwen2.5-coder-32b·0 tok·0 tools·0/20 steps·0s

How it works

One prompt. Many specialists.

Santra's orchestrator agent reads your request, explores the codebase, and delegates to focused sub-agents when the task requires it. Each specialist has a bounded role.

orchestrator

Reads your request, explores the repo, delegates or acts directly.

file-picker

Locates relevant files using glob and ripgrep search.

reader

Synthesises architecture and implementation context.

executor

Makes precise edits — str_replace or full file writes.

reviewer

Critiques diffs and flags risks before applying.

thinker

Reasons through hard decisions without using tools.

Features

Built for the terminal.

file-awareness

Reads your whole repo

The agent indexes your directory tree, reads relevant files, and understands your architecture before touching anything. Files larger than 40 KB are truncated automatically.

change-approval

Approves diffs before applying

Every file change surfaces a diff — line counts, changed content — and waits for your approval. You can accept, reject, or give feedback inline.

shell-execution

Runs terminal commands

Agents can run shell commands with configurable timeouts. Tests, linters, build steps — anything you'd run yourself.

session-memory

Persistent sessions

Conversations are auto-saved. Use /resume to reopen any past session. /copy exports the transcript to clipboard.

interactive

Agents ask you questions

When the agent needs clarification, it pauses and presents a question — with optional choices. Your answer feeds back into the run.

open-source

Extend with custom agents

Drop agent files into .agents/ to register custom specialists. Each agent defines its own system prompt, tools, and delegation rules.

Provider-agnostic

Your key. Any model.

Set one environment variable. Santra routes directly to your provider — no middleman, no markup. Pay your provider, not a subscription.

# Anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# Nvidia NIM (OpenAI-compatible)
export OPENAI_BASE_URL=https://integrate.api.nvidia.com/v1
export OPENAI_API_KEY=nvapi-...

# Ollama (local — no API key needed)
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_API_KEY=ollama

AnthropicOpenAINvidia NIMGroqTogether AIFireworksOllamaLM StudiovLLMany OpenAI-compatible API

Full provider guide →

Tooling

23 built-in tools.

Agents have access to a full toolkit — file reads, targeted edits, shell execution, web search, and structured inter-agent messaging.

File ops

read_filewrite_filestr_replaceapply_patch

search_filessearch_textcode_searchglobread_subtree

Agent control

spawn_agentspawn_agentstask_completedset_output

Interactive

ask_userweb_searchread_docsrun_terminal_command

Get started.

Install globally, set your key, and run in any repository. No account required.

$ npm install -g santra-cli
$ export ANTHROPIC_API_KEY=sk-ant-...
$ cd my-project && santra

Full install guide →Read the docs