# connic

> connic is a German, code-first AI agent deployment and orchestration platform for developers. Teams define agents in YAML, extend them with Python, and run them in a managed runtime with first-party connectors, observability, evals, memory, vector storage, BYOK, and EU-hosted connic-managed models.

## About

- **Type**: Managed runtime + platform (SaaS, with enterprise self-hosting)
- **Category**: AI agent deployment platform
- **Primary language**: Python (YAML for agent config)
- **Pricing model**: Subscription-as-credit with prepaid top-ups and optional capped auto-refill for standard plans. Purchased credit remains available while the Project is active. Enterprise contracts have custom pricing and commercial terms.
- **Plans**: Basic is €0/month with €20 free credit for the account's first Project, valid for 14 days. Basic Projects can add prepaid credit at any time without a paid plan. Developer starts at €40/month and pro at €200/month; both paid plans provide the selected monthly Project credit with 1-, 12-, or 24-month terms. Enterprise has custom pricing and commercial terms. Base platform rates before run-package discounts: €0.0337 per LLM-agent start or completed judge evaluation, €0.00042 per runtime second, €0.00342 per GB-hour, and €0.10 per 1M retrieval characters. Each successful web search costs €0.015 and each extracted page costs €0.015. connic-managed model prices are listed separately in the model catalog.
- **Discounts**: Prepaid 12-month terms save 10%; 24-month terms save 25%. Monthly credit is unchanged. Run packages reduce LLM-agent starts, completed judge evaluations, runtime, storage and retrieval by 5% from 10,000 runs, 10% from 25,000 runs, 15% from 50,000 runs, 20% from 100,000 runs, 25% from 150,000 runs, 30% from 250,000 runs, 35% from 500,000 runs. These volume discounts also apply to top-up-funded usage; model tokens and web tools keep their unit rates. See [pricing](https://connic.co/pricing.md) for the full calculation.
- **Currency**: EUR
- **Started**: 2025 in Munich, Germany
- **Company page**: [About connic](https://connic.co/about)
- **Last updated**: September 2026

## Core Product

- [Quickstart Guide](https://connic.co/docs/v1/quickstart): Create a Project, deploy an agent, and trigger a run from a connector
- [Coding Agent Setup](https://connic.co/docs/v1/ai-agent-setup): Install the connic plugin, skill, and MCP connection in Codex, Claude Code, Cursor, OpenCode, or another compatible client
- [Deployment](https://connic.co/docs/v1/platform/deployment): Build and release agents from Git or the CLI
- [Project Structure](https://connic.co/docs/v1/build/project-structure): Standard directories for agents, tools, tests, middleware, and hooks
- [Configure an Agent](https://connic.co/docs/v1/build/agent-configuration): Guided setup for choosing an agent type, model, prompt, and tools
- [connic model catalog](https://connic.co/docs/v1/build/connic-models): Exact connic/* model IDs, EUR token prices, input types, and context limits
- [Agent YAML](https://connic.co/docs/v1/build/agent-configuration/agent-yaml): Complete field reference for agent configuration files
- [connic voice](https://connic.co/docs/v1/build/voice): Native voice models using the familiar connic agent configuration, including prompts, tools, and voice settings
- [Twilio Messaging](https://connic.co/docs/v1/connectors/twilio-messaging): Receive and send SMS/MMS, WhatsApp, and RCS messages using a reusable Twilio connection
- [Twilio Voice](https://connic.co/docs/v1/connectors/twilio-voice): Connect an existing Twilio number for incoming calls to a voice agent
- [Telnyx Voice](https://connic.co/docs/v1/connectors/telnyx-voice): Connect an existing Telnyx number to a voice agent with automatic Call Control setup
- [Telnyx Messaging](https://connic.co/docs/v1/connectors/telnyx-messaging): Receive and send SMS/MMS, WhatsApp, and RCS messages through a saved Telnyx connection
- [Write the First Test](https://connic.co/docs/v1/test): Start a YAML test suite, run it, and use it as a deployment gate
- [Connector Modes](https://connic.co/docs/v1/connectors): Inbound, outbound, and synchronous integration patterns
- [Runs & Traces](https://connic.co/docs/v1/platform/observability/runs-and-traces): Inspect executions, tool calls, inputs, outputs, and context
- [Migrate from LangChain](https://connic.co/docs/v1/migrate-from-langchain): Step-by-step guide for moving LangChain projects to a managed runtime
- [REST API](https://connic.co/docs/v1/reference/rest-api): Programmatic access to agents, runs, and platform resources

## Developer and Agent Resources

- [connic MCP endpoint](https://mcp.connic.co/mcp): Authenticated Streamable HTTP endpoint for permission-scoped access to connic projects.
- [Official MCP Registry listing](https://registry.modelcontextprotocol.io/v0.1/servers/co.connic%2Fmcp/versions/latest): Versioned registry metadata for the official connic remote MCP server.
- [MCP protected-resource metadata](https://mcp.connic.co/.well-known/oauth-protected-resource/mcp): RFC 9728 discovery document for the connic MCP resource and its supported scopes.
- [MCP authorization-server metadata](https://mcp.connic.co/.well-known/oauth-authorization-server): OAuth issuer, authorization, registration, token, revocation, and JWKS endpoints for MCP clients.
- [Authentication guide](https://connic.co/auth.md): REST API-key and MCP OAuth requirements, permissions, revocation, and error recovery for agents and API clients.
- [OpenAPI specification](https://api.connic.co/v1/public/openapi.json): Machine-readable OpenAPI 3.1 schema for the public connic REST API.
- [Python SDK and CLI](https://pypi.org/project/connic-composer-sdk/): Official PyPI package; installing it provides the `connic` command-line tool.
- [connic skill](https://github.com/connic-org/connic-skill): Build, test, deploy, and operate connic agents with current composer SDK, CLI, connector, platform, and connic MCP guidance.
- [Dedicated test environments](https://connic.co/docs/v1/test/running-tests#set-up-a-dedicated-test-environment): Run suites against an isolated environment with test credentials, resources, connectors, and explicit cleanup.

## Use Cases

- [Code-First EU AI Agent Platform](https://connic.co/use-cases/code-first-eu-ai-agent-platform): Build, orchestrate, test, deploy, and operate agents with YAML, Python, Git, EU-hosted managed models, and one managed platform
- [Add AI to a Product](https://connic.co/use-cases/add-ai-to-product): Embed agents into existing SaaS products through first-party connectors
- [Replace Self-Hosting](https://connic.co/use-cases/replace-self-hosting): Keep agent code while replacing Kubernetes, deployment, scaling, and observability infrastructure

## Pricing

- [Pricing](https://connic.co/pricing): Basic at €0, developer from €40/month, pro from €200/month, selectable monthly Project credit and terms, prepaid top-ups, published usage rates, and custom enterprise terms
- [Machine-readable pricing](https://connic.co/pricing.md): Structured plan data for AI agents, including tier prices, credits, limits, and usage rates

## Features

- [Models](https://connic.co/features/models): BYOK and EU-hosted connic/* models with Project credits and published rates
- [Managed Runtime](https://connic.co/features/managed-runtime): Git-based deployment, isolated containers, autoscaling, regions, rollbacks, and runtime controls without operating Kubernetes
- [Apps](https://connic.co/features/apps): Host the frontend of your agent system with sign-in, data endpoints guarded by access rules, and per-app stats on pro and enterprise
- [Connectors](https://connic.co/features/connectors): First-party connectors for Kafka, SQS, Stripe, Postgres, Email, Telegram, Webhooks, and Cron without custom consumers or webhook handlers
- [Guardrails](https://connic.co/features/guardrails): Built-in and custom checks for prompt injection, PII, moderation, leakage, and policy enforcement before or after model calls
- [Observability](https://connic.co/features/observability): Tracing, run history, token usage, and custom dashboards included on every paid plan
- [Storage](https://connic.co/features/storage): Managed semantic retrieval, persistent sessions, and a schemaless database
- [Retrieval](https://connic.co/features/retrieval): Managed ingestion, source sync, hierarchical namespaces, semantic search, citations, and purpose-specific retrieval tools
- [AI Governance](https://connic.co/features/ai-governance): Enterprise tooling for registering AI systems, running preliminary provider/deployer assessments, creating controls, recording customer-managed Article 50 policies and attestations, tracking incidents, and exporting immutable metadata-only evidence; connic does not deliver disclosures, and the tooling is not legal advice
- [Infrastructure](https://connic.co/features/environments): Git-based deploys, auto-scaling, and isolated environments within each plan's limit

## Marketplace

- [Marketplace](https://connic.co/marketplace): Agent templates, connectors, and retrieval sources, all published by connic
- [Connectors Catalog](https://connic.co/marketplace/connectors): All first-party connectors (Kafka, SQS, Stripe, Postgres, Email, Telegram, Webhooks, Cron, WebSocket, MCP, S3, bridge)
- [Agent Templates](https://connic.co/marketplace/agent-templates): Production-ready agent starters installed via the connic CLI
- [Retrieval Sources](https://connic.co/marketplace/retrieval-sources): Sources that sync Notion, Confluence, Superhuman Docs (Coda), or any website into an agent's retrieval

## Comparisons

- [Compare connic](https://connic.co/vs): Browse all current connic platform and framework comparisons.
- [connic vs Amazon Bedrock AgentCore](https://connic.co/vs/aws-bedrock-agentcore): AWS's managed, framework-agnostic agent infrastructure spans runtime, memory, identity, gateways, observability, policy, and evaluations across multiple EU regions. connic offers a more opinionated Python and YAML path with first-party connectors and an integrated enterprise governance surface. Last reviewed 2026-10-14.
- [connic vs Google ADK + Agent Platform](https://connic.co/vs/google-agent-platform): Google's current stack combines the open-source Agent Development Kit with a managed Agent Runtime, custom containers, sessions, memory, evaluation, observability, and Google Cloud controls. connic narrows that surface into a Python and YAML operating model with managed connectors, selectable EU Project regions, and enterprise governance. Last reviewed 2026-10-14.
- [connic vs Mistral AI Studio](https://connic.co/vs/mistral-ai-studio): Mistral's Studio platform combines Agents with Workflows and Connectors (Public Preview), guardrails, and observability and evaluations in Private Preview for Enterprise, in hosted, hybrid, dedicated or self-hosted deployments. connic stays model-agnostic and adds event connectors, live A/B tests and a YAML and Python workflow. Last reviewed 2026-10-14.
- [connic vs Cloudflare Agents](https://connic.co/vs/cloudflare-agents): Cloudflare's TypeScript-native Agents runtime uses Workers and Durable Objects for durable identity, SQLite state, real-time connections, and scheduling; teams can add Workflows, AI Gateway, approval patterns, and tracing. connic offers Python and YAML authoring with named infrastructure connectors, integrated evaluation, and enterprise governance workflows. Last reviewed 2026-10-14.
- [connic vs LangSmith Deployment](https://connic.co/vs/langsmith-deployment): Framework-agnostic managed Agent Server with tracing, evaluations, an EU cloud region, and Enterprise hybrid or self-hosted deployment. connic adds a German contract, managed connectors, traffic-split testing, and integrated governance controls. Last reviewed 2026-10-14.
- [connic vs Vercel](https://connic.co/vs/vercel): TypeScript agent infrastructure built around the AI SDK, eve, AI Gateway, Workflows, Sandbox, and regional Queues. Compute, workflow state, queue persistence, and model inference can use EU regions, while end-to-end residency still depends on gateway processing, providers, configured services, and failover behavior. Last reviewed 2026-10-14.
- [connic vs n8n](https://connic.co/vs/n8n): German workflow-automation platform with LangChain-based agent nodes, evaluations, guardrails, and approvals. It is strongest at visual process automation; connic treats the agent as the versioned, deployed, and metered unit. Last reviewed 2026-10-14.
- [connic vs Mastra](https://connic.co/vs/mastra): TypeScript agent framework with self-hosting and a managed platform that can pin the server, hosted database, and observability data to an EU region. connic provides a Python and YAML operating model with managed connectors and governance controls. Last reviewed 2026-10-14.
- [connic vs Inngest + AgentKit](https://connic.co/vs/inngest-agentkit): Durable-execution platform with a JavaScript and TypeScript agent framework, evals, experiments, and human-in-the-loop workflow patterns. Managed Cloud database data is US-hosted; self-hosting is the documented residency alternative. Last reviewed 2026-10-14.
- [connic vs Agentuity](https://connic.co/vs/agentuity): Agentuity, a managed platform for AI agents, was discontinued in 2026. This page maps its agents, triggers, storage, evals and hosting to connic for teams moving their agents. Last reviewed 2026-10-14.
- [connic vs Trigger.dev](https://connic.co/vs/trigger-dev): Open-source, Git-first background-job platform with AI tasks, HITL patterns, Realtime, and LLM usage metrics. EU run execution does not relocate payloads or logs; complete residency requires self-hosting or a confirmed private deployment. Last reviewed 2026-10-14.
- [connic vs LangChain](https://connic.co/vs/langchain): MIT-licensed Python and JavaScript framework with more than 1,000 integrations. The framework remains self-hosted code, while the vendor's LangSmith platform adds managed deployment, tracing, and EU-region options under a US contract. Last reviewed 2026-10-14.
- [connic vs AutoGen + Microsoft Agent Framework](https://connic.co/vs/autogen): AutoGen is in maintenance mode, with Microsoft Agent Framework as its successor. The successor can run as self-hosted SDK code or through Foundry Hosted Agents, where residency depends on the selected resources, models, and tools. Last reviewed 2026-10-14.
- [connic vs CrewAI](https://connic.co/vs/crewai): MIT-licensed Python framework with JSONC or YAML agent files, plus CrewAI AMP for managed deployment, tracing, evaluations, integrations, knowledge, and approvals. Standard managed-region choice is not publicly specified; Enterprise offers vendor cloud, customer VPC, or customer infrastructure. Last reviewed 2026-10-14.
- [connic vs Self-Hosting](https://connic.co/vs/self-hosting): Operating an agent stack on infrastructure controlled by the organization gives direct control over the runtime and its data, subject to configured cloud, model, and tool dependencies. This comparison uses production Kubernetes as a representative baseline and counts the operating work left with the organization. Last reviewed 2026-10-14.
- [connic vs AI by Zapier](https://connic.co/vs/zapier-ai): Zapier is moving standalone Agents into AI by Zapier, an agentic step inside the Zap editor with model choice, BYOK, tools, knowledge, approvals, and task-based billing. It suits UI-driven automation; connic remains Git-native and keeps its managed core in the EU. Last reviewed 2026-10-14.

## Featured Articles

- [EU-Hosted AI Models in 2026](https://connic.co/blog/eu-hosted-ai-models-2026): connic, Mistral, Scaleway, OVHcloud, and self-hosted options compared by inference location, retention, model choice, and operating model
- [How to Deploy a Python AI Agent Without Kubernetes](https://connic.co/blog/deploy-python-ai-agent-without-kubernetes): Complete YAML, plain-Python, test-gated, Git-deployed example with an EU-hosted managed model and transparent cost calculation
- [Introducing connic models](https://connic.co/blog/introducing-connic-models): EU-hosted models built into every Project, named in one line of agent YAML, with no provider account and no API key
- [Introducing the connic marketplace](https://connic.co/blog/introducing-connic-marketplace): The new home for agent templates, connectors, and retrieval sources: one catalog, one publisher model, one CLI install
- [Best AI Agent Platforms for EU Enterprises in 2026](https://connic.co/blog/best-ai-agent-platforms-eu-2026): Ranked shortlist evaluated on EU data residency, self-hosting, MCP tool support, BYOK, and EU AI Act readiness
- [AI Agent Deployment Platforms: 16 Vendors Compared (2026)](https://connic.co/blog/ai-agent-deployment-platforms-2026): Compare managed runtimes, workflow engines, framework deployment stacks, and frameworks by language and operating scope
- [Managed vs Self-Hosted AI Agents: TCO at 50K LLM-Agent Runs](https://connic.co/blog/managed-vs-self-hosted-ai-agents-tco): Compare three-year annualized costs at 50,000 monthly LLM-agent runs across connic, a build-and-self-host stack, and a bought-services stack
- [What Is an MCP Connector? A Practical Definition](https://connic.co/blog/what-is-an-mcp-connector): How MCP connectors relate to MCP servers and clients, when they fit, and where authenticated production deployment changes the design
- [EU AI Act Enforcement: Who Investigates and What Evidence to Keep](https://connic.co/blog/eu-ai-act-compliance-ai-agents): How the AI Office, national authorities, and EDPS divide enforcement by system and provider, and which governance and runtime evidence teams should retain
- [Hidden Costs of Self-Hosting AI Agents](https://connic.co/blog/hidden-costs-self-hosting-ai-agents): Real cost breakdown covering infrastructure, DevOps time, on-call, and opportunity cost
- [AI Agents: From Prototype to Production](https://connic.co/blog/ai-agents-prototype-to-production): What changes when an agent moves from demo to serving real users
- [Migrate from LangChain to Production AI Agents](https://connic.co/blog/migrate-langchain-to-production): How to move an existing LangChain project to a managed runtime without rewriting
- [Secure AI Agents: A Production Safety Checklist](https://connic.co/blog/secure-ai-agents-production): Security checklist for AI agents in production
- [Agent Observability: Track Costs, Tokens & Runs](https://connic.co/blog/ai-agent-observability-guide): Monitoring and debugging production agents

## Legal & Compliance

- [Security](https://connic.co/legal/security): Encryption, access controls, and security practices
- [EU AI Act Compliance](https://connic.co/legal/eu-ai-act): How connic maps to EU AI Act requirements
- [AI Governance (product docs)](https://connic.co/docs/v1/platform/ai-governance): Enterprise EU AI Act readiness documentation workspace to register AI systems, run preliminary provider/deployer assessments, generate controls, record customer-managed Article 50 policies and attestations, track incidents, and export immutable metadata-only evidence; connic does not deliver disclosures and the workspace is not legal advice