When Simplicity Becomes Your Superpower: Meet KISS Multi-Agent Multi-Optimization Framework
"Everything should be made as simple as possible, but not simpler." — Albert Einstein
KISS stands for "Keep it Simple, Stupid" which is a well-known software engineering principle.
Table of Contents
Introduction to KISS Sorcar
KISS Sorcar (named after the famous Bengali magician P.C. Sorcar) is a general-purpose agent with web browsing and native desktop app execution capabilities*. It runs locally as a VS Code extension. It codes really well and works pretty fast. The agent can run relentlessly for hours to days. It is implemented as a Visual Studio Code extension. It has full browser support, multimodal support, and OpenClaw like features (we will post more information about them on our social media channels). The good part is that KISS Sorcar is completely free and open-source; you bring your API key and give it to KISS Sorcar. Note that Sorcar also means government in Bengali.
git clone https://github.com/ksenxx/kiss_ai.git
cd kiss_ai
./install.sh
https://www.youtube.com/watch?v=xnYxWvRqACE.
Install and Launch KISS Sorcar
To Install KISS Sorcar, open Visual Studio Code, search for "KISS Sorcar" in the extension marketplace, install, and relaunch VS Code. Press ESC if you don't have a specific API key, but you must provide at least one API key.
You can also manually download the extension from src/kiss/agents/vscode/kiss-sorcar.vsix.
If you do not want to use the KISS Sorcar IDE, you can open a terminal and use sorcar as a normal shell command. Some examples are:
sorcar -t "What is 2435*234"
sorcar -n -t --use-chat "What is 2435*234?" # to start in a new chat session in sorcar use -n
sorcar -m "claude-sonnet-4-6" -t "What is 2435*234?" # to use a specific model
echo "Can you find the cheapest non-stop flight from SFO to JFK on June 15 by consulting various websites?" > prompt
sorcar -f prompt # use contents of a file to send task
sorcar -t 'Can you send the message "Hello from Sorcar!" to ksen via the desktop slack app?'
sorcar -t 'Can you write a thorough and precise plan in PLAN.md to simplify the project code?'
sorcar -t 'I see some issues and bugs in PLAN.md. Can you fix them?' # lie to the agent to force improve the plan
🤖 Models Supported
Supported Models: The framework includes context length, pricing, and capability flags for:
Generation Models (text generation with function calling support):
- OpenAI: gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, gpt-4o, gpt-4o-mini, gpt-4.5-preview, gpt-4-turbo, gpt-4, gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, gpt-5.1, gpt-5.2, gpt-5.2-pro, gpt-5.3-chat-latest, gpt-5.4, gpt-5.4-pro
- OpenAI (Codex): gpt-5-codex, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini, gpt-5.2-codex, gpt-5.3-codex, codex-mini-latest
- OpenAI (Reasoning): o1, o1-mini, o1-pro, o3, o3-mini, o3-mini-high, o3-pro, o3-deep-research, o4-mini, o4-mini-high, o4-mini-deep-research
- OpenAI (Open Source): openai/gpt-oss-20b, openai/gpt-oss-120b
- Anthropic: claude-opus-4-6, claude-opus-4-5, claude-opus-4-1, claude-opus-4, claude-sonnet-4-6, claude-sonnet-4-5, claude-sonnet-4, claude-haiku-4-5
- Anthropic (Legacy): claude-3-5-haiku, claude-3-haiku-20240307
- Gemini: gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash, gemini-2.0-flash-lite, gemini-1.5-pro (deprecated), gemini-1.5-flash (deprecated)
- Gemini (preview, unreliable function calling): gemini-3-pro-preview, gemini-3-flash-preview, gemini-3.1-pro-preview, gemini-3.1-flash-lite-preview, gemini-2.5-flash-lite
- Together AI (Llama): Llama-4-Scout/Maverick (with function calling), Llama-3.x series (generation only)
- Together AI (Qwen): Qwen2.5-72B/7B-Instruct-Turbo, Qwen2.5-Coder-32B, Qwen2.5-VL-72B, Qwen3-235B series, Qwen3-Coder-480B, Qwen3-Coder-Next, Qwen3-Next-80B, Qwen3-VL-32B/8B, QwQ-32B (with function calling)
- Together AI (DeepSeek): DeepSeek-R1, DeepSeek-V3-0324, DeepSeek-V3.1 (with function calling)
- Together AI (Kimi/Moonshot): Kimi-K2-Instruct, Kimi-K2-Instruct-0905, Kimi-K2-Thinking, Kimi-K2.5
- Together AI (Mistral): Ministral-3-14B, Mistral-7B-v0.2/v0.3, Mistral-Small-24B
- Together AI (Z.AI): GLM-5.0, GLM-4.5-Air, GLM-4.7
- Together AI (Other): Nemotron-Nano-9B, Arcee (Coder-Large, Maestro-Reasoning, Virtuoso-Large, trinity-mini), DeepCogito (cogito-v2 series), google/gemma-2b/3n, Refuel-LLM-2/2-Small, essentialai/rnj-1, marin-community/marin-8b
- OpenRouter: Access to 300+ models from 60+ providers via unified API:
- OpenAI (gpt-3.5-turbo, gpt-4, gpt-4-turbo, gpt-4.1, gpt-4o variants, gpt-5/5.1/5.2/5.3/5.4 and codex variants, o1, o3, o3-pro, o4-mini, codex-mini, gpt-oss, gpt-audio)
- Anthropic (claude-3-haiku, claude-3.5-haiku/sonnet, claude-3.7-sonnet, claude-sonnet-4/4.5, claude-haiku-4.5, claude-opus-4/4.1/4.5/4.6 with 1M context)
- Google (gemini-2.0-flash, gemini-2.5-flash/pro, gemini-3-flash/pro-preview, gemma-2-9b/27b, gemma-3-4b/12b/27b, gemma-3n-e4b)
- Meta Llama (llama-3-8b/70b, llama-3.1-8b/70b/405b, llama-3.2-1b/3b/11b-vision, llama-3.3-70b, llama-4-maverick/scout, llama-guard-2/3/4)
- DeepSeek (deepseek-chat/v3/v3.1/v3.2/v3.2-speciale, deepseek-r1/r1-0528/r1-turbo, deepseek-r1-distill variants, deepseek-coder-v2, deepseek-prover-v2)
- Qwen (qwen-2.5-7b/72b, qwen-turbo/plus/max, qwen3-8b/14b/30b/32b/235b, qwen3-coder/coder-plus/coder-next/coder-flash/coder-30b, qwen3-vl variants, qwq-32b, qwen3-next-80b, qwen3-max/max-thinking)
- Amazon Nova (nova-micro/lite/pro, nova-2-lite, nova-premier)
- Cohere (command-r, command-r-plus, command-a, command-r7b)
- X.AI Grok (grok-3/3-mini/3-beta/3-mini-beta, grok-4/4-fast, grok-4.1-fast, grok-code-fast-1)
- MiniMax (minimax-01, minimax-m1, minimax-m2/m2.1/m2.5/m2-her)
- ByteDance Seed (seed-1.6, seed-1.6-flash, seed-2.0, seed-2.0-thinking)
- MoonshotAI (kimi-k2, kimi-k2-thinking, kimi-k2.5, kimi-dev-72b)
- Mistral (codestral, devstral/devstral-medium/devstral-small, mistral-large/medium/small, mixtral-8x7b/8x22b, ministral-3b/8b/14b, pixtral, voxtral)
- NVIDIA (llama-3.1-nemotron-70b/ultra-253b, llama-3.3-nemotron-super-49b, nemotron-nano-9b-v2/12b-v2-vl, nemotron-3-nano-30b)
- Z.AI/GLM (glm-5, glm-4-32b, glm-4.5/4.5-air/4.5v, glm-4.6/4.6v, glm-4.7/4.7-flash)
- AllenAI (olmo-2/3-7b/32b-instruct/think, olmo-3.1-32b-instruct/think, molmo-2-8b)
- Perplexity (sonar, sonar-pro, sonar-pro-search, sonar-deep-research, sonar-reasoning-pro)
- NousResearch (hermes-2-pro/3/4-llama series, hermes-4-70b/405b, deephermes-3)
- Baidu ERNIE (ernie-4.5 series including VL and thinking variants)
- Aurora (openrouter/aurora-alpha — free cloaked reasoning model)
- And 30+ more providers (ai21, aion-labs, alfredpros, alpindale, anthracite-org, arcee-ai, bytedance, deepcogito, essentialai, ibm-granite, inception, inflection, kwaipilot, liquid, meituan, morph, nex-agi, opengvlab, prime-intellect, relace, sao10k, stepfun-ai, tencent, thedrummer, tngtech, upstage, writer, xiaomi, etc.)
Embedding Models (for RAG and semantic search):
- OpenAI: text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002
- Google: text-embedding-004, gemini-embedding-001
- Together AI: BAAI/bge-large-en-v1.5, BAAI/bge-base-en-v1.5, m2-bert-80M-32k-retrieval, multilingual-e5-large-instruct, gte-modernbert-base
Each model in MODEL_INFO includes capability flags:
is_function_calling_supported: Whether the model reliably supports tool/function callingis_generation_supported: Whether the model supports text generationis_embedding_supported: Whether the model is an embedding model
🤗 Contributing
Contributions in the form of issues are welcome! KISS Sorcar should be able to take care of them.
📄 License
Apache-2.0
✍️ Authors
| |
