██████ █████ ████ ██ ██ ██ ██ ████ ██ ██ ██ ██ ██ ███ ██ ██ ██ ██ ████ █████ ██████ ██████ ████ ██ ██ ██ ██ ██ ██ ██ ███ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████
Coding help, pentesting, SOC analysis — powered by any AI provider. Bring your own keys. Works offline with Ollama.
pipx install franki-cli
franki lives in your terminal, talks to multiple AI providers, and falls back automatically when one hits a rate limit.
Groq, Gemini, OpenRouter, Anthropic, Ollama, Mistral, Cerebras, Cohere, Together, Azure and more. Automatic fallback when rate-limited.
Reads, writes, and edits files. Runs shell commands. Searches the web. Applies patches. All with your approval.
Switch to /skill pentest, /skill soc, or /skill security for MITRE mapping, payload generation, CEH/OSCP prep, and alert triage.
Point it at a local Ollama instance. No API key needed. Type /ollama to pick a model from what's installed.
Automatically fetches live results when your message contains latest, CVE-, or news. Powered by Tavily or DuckDuckGo.
Use @file.py, @src/, or @git to inject files, directories, or git state directly into your message.
Remembers facts, your pentest scope, and notes across sessions with /remember and /memories.
Connect any MCP-compatible tool server. Manage connections with /mcp and they become part of the agent's toolset.
verify_token(). The token payload's exp field was never compared against datetime.utcnow(). Also updated the 401 response to distinguish expired vs invalid tokens.
franki is fully documented, from a one-minute install to deep internals for contributors.
Install franki in 3 steps with pipx. Add your API keys, run the setup wizard, and start chatting. Works with free-tier providers — no credit card needed.
Install guide →11 providers supported — Groq, Gemini, OpenRouter, Ollama, Anthropic, Mistral, Cerebras, Together AI, Cohere, Azure, and any custom OpenAI-compatible endpoint. Automatic fallback when one hits a rate limit.
Browse providers →Every slash command documented — conversation flow, agent tools, skills, long-term memory, security tools, session branching, MCP servers, one-shot CLI and more.
View docs →Full architectural overview, the 10-step request lifecycle, routing internals, and step-by-step recipes for adding providers, slash commands, skills, and agent tools.
Read the guide →