v0.1.5

• API errors (401/400/404/429) now show a helpful message instead of crashing with a stack trace
• REPL stays alive after an API error — you can fix config and keep going

v0.1.4

• Add --version / -v flag
• Show version in REPL header and --help output
• Friendly error when no API key configured (from v0.1.3)

v0.1.3

• Show friendly setup instructions when no API key is configured (instead of crashing with a raw 401)

v0.1.2

New: picante config commands

# Set active provider + store API key
picante config provider opencode "sk-..."
picante config provider openrouter "sk-or-..."

# Interactive model picker (lists from current provider)
picante config model

# Set model directly
picante config model glm-5.3-flash

# Show current settings
picante config show

Multiple provider keys are stored side-by-side in ~/.picante/config.toml — switch between providers any time with config provider.

v0.1.1

• Fix --help / -h crashing without config (was treated as a prompt)
• Config is now loaded only when the agent actually runs
• Fix --resume <id> not passing session ID correctly

v0.1.0 — initial release

First release of picante — a terminal-native AI agent built with Bun.

Install

curl -fsSL https://picanteverde.github.io/picante/install.sh | bash

What's included

• OpenAI-compatible LLM backend (any provider)
• Built-in tools: read/write files, run shell, run monitor, web search, web browse, web download
• Provider plugins: openrouter, opencode.ai, fal.ai, NVIDIA, AMD
• Sessions persisted to ~/.picante/sessions/
• Skills system via Markdown files in ~/.picante/skills/
• Auto-generates x-opencode-session + User-Agent: picante/1.0 for opencode.ai requests