# Kopern — AI Agent Builder, Orchestrator & Grader

> The open-source platform to build, grade, optimize, and deploy AI agents — no code required.

## What Kopern Does

Kopern is a web platform for creating production-grade AI agents. Describe what you want in plain text — Kopern generates the agent, tests it against 6 quality criteria, optimizes it automatically, and deploys it in one click across 5 channels. 130+ features shipped.

## Core Capabilities

- **Zero-Code Agent Builder**: AI wizard generates complete agents (system prompt, skills, tools, grading suite) from a text description. 37 ready-to-deploy templates (28 general + 9 vertical business).
- **Deterministic Grading**: 6 criterion types (output match, schema validation, tool usage, safety check, custom script, LLM judge). Scheduled grading via Vercel Cron with score drop alerts.
- **Optimization Lab (6 modes)**: AutoTune, AutoFix, Stress Lab (adversarial red team), Tournament, Distillation, Evolution — agents self-improve automatically.
- **Visual Multi-Agent Orchestration**: React Flow v12 editor with 5 node types (Agent, Condition, Trigger, Output, Export). Drag-and-drop, double-click editing, runtime status.
- **Deploy Everywhere**: Embeddable widget, Slack bot, Telegram bot, WhatsApp (Meta Cloud API), Webhooks (n8n/Zapier/Make compatible), MCP protocol (Claude Code/Cursor).
- **Built-in Agent Tools**: web_fetch (HTTP), code_interpreter (Python/Node/Bash via GCP Cloud Run), agent memory (persistent KV), email (Gmail+Outlook), calendar (Google+Microsoft), GitHub integration.
- **Agent Memory**: Persistent key-value memory across sessions with LRU eviction, cross-session search, context compaction via Haiku summarization.
- **Service Connectors**: OAuth Google (Gmail + Calendar) and Microsoft (Outlook + Graph). 8 agent tools with daily limits and tool approval.
- **Workflow Quality Monitor**: 18 standardized prompts across 6 criteria test any LLM model or API endpoint for reasoning depth, instruction following, consistency, latency, edge cases, and output quality. Compare against baselines for 29 models across 4 providers.
- **Public Agent Grader**: Free public grader at kopern.ai/grader — grade any AI endpoint or system prompt against custom test cases with 6 evaluation criteria.
- **Social Media Integration**: Bluesky AT Protocol connector with 9 tools for social media management (create posts, threads, read feeds, track metrics, search mentions, reply, delete, profiles, list connected).
- **Blog**: Technical blog at kopern.ai/blog covering AI agent development, grading systems, and monitoring.
- **EU AI Act Compliance**: Automated compliance reports (Art. 6, 12, 14, 52), tool approval for destructive actions, conversational approval on messaging channels.
- **Stripe Billing**: 4 tiers (Starter free, Pro $79/mo, Usage PAYG, Enterprise $499/mo) with usage-based meters.

## LLM Support

Multi-provider with native tool calling and streaming:
- Anthropic Claude (Opus 4.6, Sonnet 4.6, Haiku 4.5)
- OpenAI GPT (GPT-5.x, GPT-4o, GPT-4.1, o3/o4)
- Google Gemini (2.5 Pro, 2.5 Flash, 2.0 Flash)
- Mistral AI (Mistral Large, Codestral, Devstral, Magistral)
- Ollama (any local model)

Key rotation failover: up to 5 keys per provider with automatic retry on 429.

## MCP Protocol

Any Kopern agent works as an MCP server for Claude Code, Cursor, and any MCP client:

```json
{
  "mcpServers": {
    "my-agent": {
      "type": "http",
      "url": "https://kopern.ai/api/mcp/server",
      "headers": { "Authorization": "Bearer kpn_..." }
    }
  }
}
```

32 tools covering the full agent lifecycle:
- **Agent CRUD**: `kopern_create_agent`, `kopern_get_agent`, `kopern_update_agent`, `kopern_delete_agent`, `kopern_list_agents`
- **Templates**: `kopern_list_templates`, `kopern_deploy_template` (1-click deploy with skills, tools, grading)
- **Chat**: `kopern_chat` (agent-bound key), `kopern_agent_info` (agent-bound key)
- **Grading**: `kopern_grade_prompt` (inline), `kopern_create_grading_suite`, `kopern_run_grading`, `kopern_get_grading_results`, `kopern_list_grading_runs`
- **Optimization**: `kopern_run_autoresearch` (AutoTune iterative prompt optimization)
- **Teams & Pipelines**: `kopern_create_team`, `kopern_run_team`, `kopern_create_pipeline`, `kopern_run_pipeline`
- **Connectors**: `kopern_connect_widget`, `kopern_connect_telegram`, `kopern_connect_whatsapp`, `kopern_connect_slack`, `kopern_connect_webhook`, `kopern_connect_email`, `kopern_connect_calendar`
- **Monitoring**: `kopern_list_sessions`, `kopern_get_session`, `kopern_manage_memory`, `kopern_compliance_report`, `kopern_get_usage`
- **Portability**: `kopern_export_agent`, `kopern_import_agent`

Two key types: **user-level** (30 platform tools, no agent needed) and **agent-bound** (all 32 tools, includes chat).

## Self-Hosted

Docker deployment for enterprises with data sovereignty requirements:
- `docker compose up` with `docker-compose.yml` (Next.js + Redis + optional Ollama + optional Firebase Emulator)
- Supports Ollama (full local), Azure OpenAI, and any OpenAI-compatible endpoint for internal LLMs

## Architecture

- Next.js 16 + React 19 + TypeScript (strict mode)
- Firebase (Firestore real-time, Auth with Google/GitHub/Email)
- React Flow v12 (visual orchestration editor)
- GCP Cloud Run (code interpreter sandbox)
- Stripe Billing (subscriptions + usage meters)
- Upstash Redis (8 rate limiters)
- i18n: English and French

## Key API Endpoints

- `POST /api/agents/{id}/chat` — SSE streaming chat with tool calling
- `POST /api/agents/{id}/grading/{suiteId}/run` — Grading runner
- `POST /api/agents/{id}/autoresearch/*` — 6 optimization modes
- `POST /api/mcp/server` — MCP Streamable HTTP protocol
- `POST /api/widget/chat` — Embeddable widget (SSE + CORS)
- `POST /api/webhook/{agentId}` — Inbound webhooks (sync JSON)
- `POST /api/slack/events` — Slack Events API
- `POST /api/telegram/webhook` — Telegram Bot API
- `POST /api/whatsapp/webhook` — WhatsApp Cloud API
- `POST /api/teams/{id}/execute` — Team execution SSE
- `POST /api/monitor/run` — Workflow quality diagnostic (18 prompts, SSE)
- `POST /api/grader/run` — Public agent grader (rate-limited, no auth)
- `POST /api/grader/probe` — Endpoint probe for grader

## Links

- Website: https://kopern.ai
- GitHub: https://github.com/berch-t/kopern
- Full documentation: /llms-full.txt