Skip to main content
Connic
EU AI Act readyGDPRSecurity compliance

Deploy AI Agents
at Enterprise Scale

Define agents in YAML, write tools in Python, push to Git. We run them. Deployment, scaling, and the connectors into your stack are already wired up. You don't run Kubernetes or write glue code.

Get Started Free
0
agent runs this year
Read the documentationExplore Agent Templates
Build

Composer SDK

Define agents in YAML. Add custom tools when you need them.

Replaces
LangChain logoLlamaIndex logoVercel AI SDK logo
agents/invoice-processor.yaml
# Define the whole agent in one YAML file
name: invoice-processor
model: gemini/gemini-2.5-pro
temperature: 0.3
system_prompt: |
You are an expert accountant.
Extract every field from the invoice
and verify the totals add up.
tools:
  - documents.parse
  - documents.extract_entities
  - database.store_invoice

Tools

Predefined integrations, your own code tools, or any MCP server.

Replaces
Composio logo

Tools

6 attached
  • Custominvoices.parse_pdf
  • Custominvoices.store
  • Predefinedweb_search
  • MCPlinear.create_issue
  • MCPcontext7.search_docs
  • Custombilling.calculate_vat

Dev server

connic dev streams every save to a live runner. Test in seconds, not minutes.

Replaces
ngrok logo
zsh — connic dev
~/invoice-bot ❯ connic dev
  watching agents/ and tools/
  connected to dev runner · session a1b2c3d4
  ✓ ready in 412ms
  ↻ tools/documents.py changed
  ✓ reloaded invoice-processor in 2.4s
  → invoice-processor #a1b2c3d4
    parsed 3 entities, stored to db
  ✓ run completed in 2.3s
❯

Built-in storage

Retrieval, database, session memory. Nothing for you to provision.

Replaces
Pinecone logoSupabase logoRedis logo

Retrieval

24 entries · 3 namespaces
Search…
SourceEntry IDNamespace
  • invoice-template.pdf
    inv_a1b2c3
    policies.finance
  • tax-rules-2026.md
    tax_d4e5f6
    policies.finance
  • refund-faq.txt
    faq_g7h8i9
    support.faq
  • product-catalog.png
    cat_j0k1l2
    products
  • shipping-policy.md
    shp_m3n4o5
    support.shipping
  • vendor-contract.pdf
    vnd_p6q7r8
    policies.legal
Deploy

Testing framework

Write assertions for agent behavior. Run them in CI.

Replaces
Promptfoo logoBraintrust logo

tests/invoice-processor.yaml

4 passed1 failed
  • extracts invoice total312ms
  • amount > 04ms
  • currency in [EUR, USD]6ms
  • vat is correctly calculated18ms
  • vendor name extracted11ms
$connic test351ms total

Environments

Dev / staging / prod with isolated secrets and config.

Replaces
Doppler logoGitHub Actions environments logo

Environment variables

production
KeyValue
  • DATABASE_URLpostgres://••••@db.connic.co/prod
  • STRIPE_SECRET_KEYsk_live_••••••••••••••••••
  • OPENAI_API_KEYsk-••••••••••••••••••••••
  • SENTRY_DSNhttps://••••@o123.sentry.io/4501
  • AWS_S3_BUCKETconnic-prod-uploads
  • WEBHOOK_SIGNING_SECRETwhsec_••••••••••••••••••

Private network bridges

Reach internal services without opening firewall ports.

Replaces
ngrok logoCloudflare Tunnel logoTailscale logo

Bridges

Production VPCconnected
last seen 2s ago
Bridge ID
bridge_8f3a2c91d4e6
Run on your network
docker run -d --name connic-bridge \
  -e BRIDGE_TOKEN=••••••••••••••••
  -e ALLOWED_HOSTS=postgres:5432,kafka:9092 \
  connicorg/bridge:latest

Connectors

Webhooks, queues, databases, schedules. Wire agents to the systems you already run.

Replaces

Connectors

inbound & outboundall healthy
trigger
stripe.invoice.paid
agent
invoice-processor
Webhook
Cron
Postgres
Kafka
Stripe
Email
SQS
Telegram
WebSocket
S3
MCP
Ship

Observability

Every run traced. Tokens, latency, errors, costs.

Replaces
Langfuse logoDatadog logoHelicone logo

Run details

#a1b2c3d4
completed
2.34s1,234 tokens$0.018Dec 5 · 14:34:12
  1. Runinvoice-processor2,341msok
  2. LLMgemini-2.5-pro · 1,203 tokens1,892msok
  3. Tooldocuments.parse234msok
  4. MCPlinear.create_issue156msok
  5. Tooldatabase.store_invoice59msok

Judges

LLM-as-judge evaluation built into the platform.

Replaces
Braintrust logoPatronus logo

Judges

last 24h
StatusJudgeAvg ScoreEvaluated
  • Active
    answer_accuracy
    on support_agent
    86%
    412 runs
  • Active
    tool_usage
    on support_agent
    91%
    412 runs
  • Active
    tone_match
    on invoice-processor
    64%
    218 runs
  • Active
    completeness
    on support_agent
    89%
    412 runs
  • Active
    response_quality
    on invoice-processor
    97%
    218 runs

A/B testing

Route traffic between agent variants. Compare outcomes.

Replaces
Statsig logoGrowthBook logo

order-handler

A/B test · 4,128 runs

running
control50%
49%
success rate
test-friendly-tone50%
68%
+19pp vs control

Approvals (HITL)

Pause tool calls for human review before they execute.

Replaces
Slack approval bots logo

Approvals

Pending
order-handler
stripe.refund
2m ago
customera.smith@acme.com
amount€129.00
reasonduplicate charge

Any model. Any system. One platform.

Use your own API keys with any provider. Connect agents to the infrastructure you already run.

Models
  • OpenAI
  • Anthropic
  • Google Gemini
  • Azure OpenAI
  • AWS Bedrock
  • Vertex AI
  • OpenRouter
Connic
Connectors
  • Webhooks
  • Cron
  • Kafka
  • SQS
  • PostgreSQL
  • WebSocket
  • MCP
  • Email
  • S3
  • Stripe

Start free. Scale when you're ready.

No credit card required. Upgrade as your agents grow.

Free

€0forever

Try Connic with real agents

  • €25 credit your first month, €1/mo after
  • 3 active connectors
  • 1 environment
  • 14 day run history
Start Building

Developer

€40/ month

Ship your first production agent

  • €40 in monthly usage credit
  • 10 active connectors · 3 environments
  • 10 min run timeout
  • 30 day log retention
Start with Developer
Most Popular

Pro

€200/ month

For teams running agents in production

  • €200 in monthly usage credit
  • 40 active connectors · 5 environments
  • 30 min run timeout · custom domains
  • 90 day log retention · priority support
Start with Pro
Compare all plans in detail

Need custom limits or enterprise compliance? Talk to us

From the blog

Tutorials, product updates, and engineering insights.

Industry Insights

EU AI Act Article 50: What Your AI Agent Must Disclose

The Commission adopted its final Article 50 guidelines on 20 July 2026, thirteen days before the rules apply. Here is what agent teams have to disclose, and when.

10 min read
Industry Insights

The OpenAI Hugging Face Hack: Guardrail Lessons for AI Agents

OpenAI models escaped a test sandbox and breached Hugging Face in July 2026. What the incident reveals about guardrails and how to secure production AI agents.

9 min read
Product Spotlight

LLM Context Compression for Long-Running AI Agents

Connic compresses older conversation history and oversized tool results, then retries the model call, so long-running agent sessions survive context limits.

8 min read
View all posts

Frequently Asked Questions

What Connic is, how it works, and what it costs.

Connic runs production AI agents for you. Define an agent in YAML, write its tools as Python functions, and push to Git. Deployment, scaling, monitoring, and the connectors that wire agents into your stack happen on our side.

No. If you can write YAML and Python, you can ship a production agent on Connic. The platform handles builds, deploys, scaling, and monitoring, so there is no Docker, Kubernetes, or CI/CD pipeline to maintain.

Connectors wire your agents into the rest of your stack. Inbound connectors trigger an agent from a webhook, cron schedule, inbox, message queue (Kafka, SQS), or Postgres event. Outbound connectors push results back to your APIs, services, or S3. Sync connectors handle real-time request/response over WebSocket. You set them up in the dashboard, and Connic runs the integration layer. Read why pre-built connectors beat custom integration glue.

Most of them: OpenAI, Anthropic (Claude), Google (Gemini), Mistral, AWS Bedrock, Azure OpenAI, plus anything compatible with the OpenAI or LiteLLM API. Swapping models is one line in your agent's YAML.

Connic meters runs, compute seconds, and storage. A paid subscription is monthly usage credit at the same per-unit rates, with overage only when usage exceeds that credit. Free includes starter credit and stops when it is exhausted. No per-seat fees.

Yes. Upload documents to the built-in Retrieval, paste text, or write to it via API. Your agents query it with semantic search. You don't have to set up a vector database, an embedding pipeline, or a chunking strategy.

Every run is captured: full execution trace, tool calls, LLM prompts and responses, tokens, latency, outcome. Click a run in the dashboard and step through it. Nothing to instrument; observability is on by default.

Yes. The Free tier includes €25 credit in your first month, then €1/month credit, three active connectors, and the full SDK and dashboard. No credit card is required. Upgrade when you need more usage or higher limits.