Skip to main content
Connic

Connic Documentation

Documentation for building, evaluating, deploying, and operating AI agents with the Composer SDK and Connic platform. Task-based guides lead to the relevant configuration reference.

Last updated
On this page

Ready to deploy a first agent?

Create a Project, connect a repository, and follow the quickstart with the dashboard open alongside it.

Start free

Start with the current task

Each route starts with a working workflow and links to the relevant configuration reference.

Build a first agent

Create a project, define an agent, deploy it, and trigger the first run

Evaluate agent behavior

Write YAML tests for outputs and tool calls before deployment

Deploy with test gates

Map environments to branches, run tests, activate versions, and roll back

Operate production runs

Inspect inputs, outputs, traces, logs, latency, and usage

Use Connic with AI toolsPlugin, skill, and MCP setup

The Connic plugin adds the Connic authoring skill and an authenticated, permission-scoped MCP connection to compatible coding agents. Agents can then work with a Connic project based on a task description.

Plugin formats and installation steps vary by client. The Coding agent setup guide provides verified plugin instructions and the separate skill and MCP path for other compatible clients.

How Connic Works

Connic is a platform for building and deploying AI agents. Here is how the pieces fit together.

Application Code

Agent Configuration + Python Tools

→

Connic Composer SDK

Validates and packages agents

↓
Connic

Connic Platform

Deploys, runs, and monitors agents

Deployments

Automated builds

Execution

Scalable processing

Observability

Runs and traces

↓

Connectors

Bridge between agents and the outside world

Inbound

Trigger agents

Outbound

Deliver results

Sync

Request-response

Key Concepts

  1. Build

    Projects and environments
    A project is the top-level container for agents and platform settings. Environments isolate variables, connectors, deployments, retrieval and database data, and run history across development, staging, and production.
    Agents
    Agents are defined in YAML as LLM, sequential, or deterministic tool agents and can be combined for workflows with multiple steps.
    Tools
    Tools extend agents with built-in capabilities, typed Python functions, MCP servers, or OpenAPI specifications.

  2. Ship

    Tests and deployments
    Repeatable YAML tests check deployed agent behavior and can gate deployments. Deployments run from a mapped Git branch or the CLI; successful versions remain available for rollback.
    Connectors
    Inbound connectors trigger agents; outbound connectors deliver results, and sync connectors wait for a response. Connector configuration stays isolated by environment.
    REST API
    Project-scoped API keys provide access to manage and observe platform resources. External systems start agents through connectors.

  3. Operate

    Runs and traces
    Every execution becomes a run. Its input, output, context, logs, and trace are available alongside usage by agent and model.
    Retrieval and database
    Retrieval searches documents, while the database stores schemaless application state. Both remain isolated by environment.
    Safety and quality
    Guardrails inspect inputs and outputs; approvals pause sensitive tool calls; judges score selected runs; and A/B tests compare live variants.
Quick StartCreate and deploy a first agent
terminal
# Install the SDK
pip install connic-composer-sdk

# Create a new project
connic init my-agents
cd my-agents

# Push to the connected repository to deploy
git push origin <branch>
Read the full guide

Choose the next step

Quickstart

Create, deploy, and trigger a first agent

Migrate from LangChain

Bring an existing LangChain or LangGraph app into Connic

Migrate from ADK

Bring an existing Google ADK project into Connic

Agent Configuration

Learn how to configure LLM, Sequential, and Tool agents