No infrastructure expertise required

Build Agents
Like You Build Apps

Stop hiring infrastructure specialists. Any engineer can ship production AI agents using familiar tools: YAML configs, Python functions, and Git. No DevOps, no framework overhead.

Get Started Read the SDK docs

CLI commands

YAML per agent

DevOps needed

agents/invoice-processor.yaml

1version: "1.0"
2
3name: invoice-processor
4model: gemini/gemini-2.5-pro
5description: "Extract invoice data"
6
7system_prompt: | 
8  Extract invoice details and
9  save them to the database.
10
11tools: 
12  -  invoice.parse_pdf
13  -  database.save_invoice
14  -  notifications.send

Familiar tools, zero learning curve

If you know YAML and Python, you already know how to build agents. No frameworks to learn, no abstractions to fight.

YAML Configuration

Define agents declaratively. Model, prompt, tools: all in one readable file. No boilerplate, no framework magic.

Python Tools

Write tools as plain Python functions. Type hints become schemas. Docstrings become descriptions. That's it.

CLI Tooling

Three commands: init, dev, tools. Scaffold projects, validate configs, list tools. No complex setup.

Git-Native Workflow

Push to deploy. Review agents in PRs. Roll back with git revert. Your agents live in version control.

Learn about deployments

Environment Variables

API keys and secrets stay secure. Define per-environment variables in the dashboard. Access with os.environ.

Manage secrets

MCP Integration

Connect external MCP servers for additional tools. Extend capabilities without writing code.

Try the SDK with a free account

From code to production in three steps

No infrastructure expertise. No DevOps bottlenecks. Any engineer can deploy agents in minutes.

Define

Write your agent in YAML. Add Python tools. Use the CLI to validate.

connic init my-agent
# Edit agents/assistant.yaml
connic dev  # Validate

Push

Commit and push to your connected repository. Connic detects the change automatically.

git add .
git commit -m "Add invoice agent"
git push origin main

Done

Your agent is live. No Docker, no Kubernetes, no infra configs. It just works.

# Agent deployed! 🎉
# Trigger via connectors
# Monitor in dashboard

What you don't need

Traditional agent deployment requires DevOps expertise, infrastructure management, and complex tooling. With Connic SDK, you skip all of that.

Focus on building agents, not managing infrastructure. Let your engineers ship features instead of wrestling with deployment.

Kubernetes configs

Docker expertise

CI/CD pipelines

Infrastructure code

DevOps knowledge

Framework lock-in

YAML configuration

Python functions

Git workflow

Instant deploys

Zero infrastructure

Any engineer can ship

What you get instead

A simple workflow that any developer can follow. Define agents in YAML, write tools in Python, push to Git. Done.

No dedicated AI infrastructure role. No waiting on DevOps. Ship agents as fast as you ship features.

Explore more capabilities

Connectors

Trigger your agents

Knowledge

Give agents memory

Observability

Monitor everything

See it in action

Simple, readable code that does exactly what you expect. No magic, no hidden complexity.

agents/document-analyzer.yaml

version: "1.0"

name: document-analyzer
model: gemini/gemini-2.5-pro
description: "Analyze documents and extract insights"

system_prompt: |
  You are a document analysis expert.
  Extract key information, summarize content,
  and identify important entities.

tools:
  - documents.parse
  - documents.extract_entities
  - database.store_analysis
  - notifications.send_report

temperature: 0.3
max_concurrent_runs: 5

Build Agents Like You Build Apps