Kendr

Open-source multi-agent intelligence platform — Web UI + CLI for research, project management, and software automation.

Kendr is a Python runtime that combines specialized AI agents, a web-based chat and project UI, multi-source research, durable memory, and structured run artifacts. Use it from the browser or the terminal.

Quickstart · CLI Reference · Configuration · Integrations · Examples

Install

Requirements

• Python 3.10 or newer — python.org/downloads
• Git — git-scm.com
• An OpenAI API key (or Anthropic / Google / local Ollama — see LLM Providers below)

macOS / Linux — one command

git clone https://github.com/kendr-ai/kendr.git
cd kendr
./scripts/install.sh

After it finishes, reload your shell and run kendr --help to confirm the install:

source ~/.zshrc     # zsh (macOS default)
# or
source ~/.bashrc    # bash (Linux default)

kendr --help

Windows — PowerShell

Open PowerShell (not CMD) and run:

git clone https://github.com/kendr-ai/kendr.git
cd kendr
powershell -ExecutionPolicy Bypass -File .\scripts\install.ps1

Open a new terminal after the script finishes, then verify:

kendr --help

Manual install (any platform)

If you prefer to control the environment yourself:

git clone https://github.com/kendr-ai/kendr.git
cd kendr

# Create and activate a virtual environment
python3 -m venv .venv
source .venv/bin/activate       # macOS/Linux
# .venv\Scripts\activate        # Windows PowerShell

# Install kendr
pip install -e .

# Verify
kendr --help

Install with pip (no git clone)

If you just want the package without cloning the repo:

pip install kendr-runtime
kendr --help

Note: Running kendr ui from a pip install requires the repo to be present for the HTML assets. Clone + install is recommended for the full experience.

LLM Providers

Kendr ships with OpenAI by default. Install optional packages to add more providers:

Or use the install script with --full:

./scripts/install.sh --full          # macOS/Linux
.\scripts\install.ps1 -Full         # Windows

First-Time Setup

After installing, set two required values:

# Your LLM API key
kendr setup set openai OPENAI_API_KEY sk-...

# Where kendr writes output files
kendr setup set core_runtime KENDR_WORKING_DIR ~/kendr-work

# Check everything is configured
kendr setup status

Or copy .env.example to .env and fill in the values manually.

Launch the Web UI

kendr ui
# or the shorter alias:
kendr web

Opens the web interface at http://localhost:5000 with:

• Chat — multi-agent chat with streaming output and plan cards
• Projects — open any local code project, chat with an AI that understands your codebase, manage files, run terminal commands, view git status
• Setup & Config — configure API keys and LLM providers in the browser
• Run History — browse every past run and its output artifacts
• LLM Models — view and switch between available models

CLI Quick Reference

# Research
kendr run "Analyse the AI chip market: key players, supply chain, investment outlook."
kendr research --sources arxiv,web --pages 15 "Advances in LLM reasoning 2024"

# Software project generation
kendr generate --stack fastapi_postgres "Task management API with auth and tests."
kendr generate --stack nextjs_prisma_postgres "Blog platform with markdown and auth."

# SuperRAG knowledge sessions
kendr run --superrag-mode build --superrag-new-session --superrag-session-title "docs" \
  --superrag-path ./docs "Index my documentation."
kendr run --superrag-mode chat --superrag-session docs "What are the install requirements?"

# Shell command execution (with approval gate)
kendr run --os-command "df -h" --os-shell bash --privileged-approved "Check disk usage."

# Gateway (required for communication integrations and REST API)
kendr gateway start
kendr gateway status
kendr gateway stop

What You Can Do

Feature Status Matrix

Capability Highlights

Web UI and Project Workspace

kendr ui
# → http://localhost:5000

The web interface gives you a full project workspace:

• Open any local Git repository and chat with the AI about your code
• Auto-generates a kendr.md project context file if one doesn't exist
• Model selector and live context-window usage bar in the chat input
• File browser, integrated terminal, and git status panel
• Recent chat history per project

Deep Research and Document Generation

Multi-source research that synthesises web, academic, patent, and local evidence into structured reports.

kendr run --current-folder \
  "Create an intelligence brief on Stripe: business model, competitors, strategy, risks."

kendr run --long-document --long-document-pages 50 \
  "Produce an exhaustive global gold market investment report."

Multi-Agent Software Project Generation

Blueprint → scaffold → build → test → verify → zip export, fully automated.

kendr generate --stack fastapi_react_postgres \
  "SaaS starter with billing, admin dashboard, and CI/CD."

Available stacks: fastapi_postgres, fastapi_react_postgres, nextjs_prisma_postgres, express_prisma_postgres, mern_microservices_mongodb, pern_postgres, nextjs_static_site, django_react_postgres, custom_freeform.

SuperRAG Knowledge Engine

Zero-config vector search over local files, URLs, databases, and OneDrive content.

kendr run --superrag-mode build --superrag-new-session \
  --superrag-session-title "product_docs" --superrag-path ./docs \
  "Index our documentation."

kendr run --superrag-mode chat --superrag-session product_docs \
  --superrag-chat "What are the installation requirements?"

Docs

• Quickstart — install, configure, first run
• CLI Reference — every subcommand and flag
• Configuration — every environment variable
• Integrations — vector backends, communication providers, security tools
• Agents — workflow families and the full agent inventory
• Architecture — runtime flow, discovery, persistence
• Security — safety boundaries and privileged controls
• Examples — copy-paste CLI examples for every workflow
• Troubleshooting — common first-run issues

Contribute

• CONTRIBUTING.md
• CHANGELOG.md
• RELEASING.md
• SECURITY.md

Contribution baseline:

• keep changes grounded in real code and verified workflows
• preserve setup-aware gating and runtime behaviour unless fixing a bug
• update tests and docs with user-facing changes
• run python scripts/verify.py before opening a PR

Community

• Join the Kendr Discord community for live chat, collaboration, Q&A, and announcements: https://discord.gg/GgU8UEdn
• Track ongoing work, open issues/feedback, and CI results on the GitHub project view
• The badges above pull live GitHub metrics (stars, forks, watchers, issues, workflows + more) so you can see how the project is performing in real time.

Contributors

Names and avatars shown above are generated automatically from GitHub contributions, so the list reflects the latest community participation without manual updates.

Under the Hood

• multi-agent orchestration runtime in kendr/runtime.py
• web UI server (chat + project workspace) in kendr/ui_server.py
• CLI entrypoint in kendr/cli.py
• dynamic agent registry and discovery in kendr/discovery.py
• multi-provider LLM routing in kendr/llm_router.py
• project context management in kendr/project_context.py
• rich terminal output in kendr/cli_output.py
• setup and integration catalog in kendr/setup/
• durable SQLite persistence in kendr/persistence/
• multi-source research infrastructure in tasks/research_infra.py
• optional HTTP gateway in kendr/gateway_server.py
• MCP server endpoints in mcp_servers/