AI-first data engineering automation for Developers
Connect your AI tools over MCP, automate anything with the GraphQL API, and access your workspace from the terminal or Python.
Paradime MCP
Connect AI tools to your data stack
The Paradime MCP server gives you access to code, warehouse, catalog, and pipelines, in Claude, Cursor, ChatGPT, or Copilot.
DinoAICode, Data Warehouse, Metadata, Pipeline Logs
Connect your AI tools to Paradime
The Paradime MCP server gives Claude, ChatGPT, Cursor, and any other MCP client tools across your repos, warehouse, catalog, and Bolt schedules — with your own role and workspace permissions.
SalesforceConnectedOAuth
GongConnectedOAuth
AtlanConnectedOAuth
HubSpotConnectedOAuth
DataHubConnectedPublicBring your own MCP servers
Connect Salesforce, DataHub, Atlan — or, in the enterprise, your in-house AI gateway where every tool is already validated — and DinoAI agents and the Code IDE copilot can call their tools.
GraphQL API
Full API access to your workspace
Trigger Bolt runs, discover metadata, manage users and workspaces, read audit logs, and trigger DinoAI agents. All of it through one GraphQL endpoint, authenticated with account API keys.
mutation TriggerBoltRun($slug: String) {
triggerBoltRun(slug: $slug) {
runId
}
}mutation TriggerDinoaiAgentRun($agent: String, $message: String) {
triggerDinoaiAgentRun(agent: $agent, message: $message) {
ok
agentSessionId
status
}
}query EnvironmentOverview($environmentSlug: String!) {
environment(slug: $environmentSlug) {
dbtProjectName
adapterType
applied {
lastUpdatedAt
resourceCounts
packages
}
}
}mutation InviteUser($email: String!, $accountType: UserAccountType!) {
inviteUser(email: $email, accountType: $accountType) {
ok
}
}Model Providers
Every frontier model, one gateway
DinoAI's model picker spans six providers. Pick a fast model for quick questions and everyday edits, or a frontier model for the hard refactor. Every request runs through Paradime's LLM gateway, so sessions and tools work the same whichever model you pick.
AnthropicClaude Opus 5, Sonnet 5, Haiku 4.5
OpenAIGPT-5.6 Sol, Terra, and Luna
GoogleGemini 3.7 Flash
xAIGrok 4.6
Moonshot AIKimi K2 Thinking
QwenQwen3 235BModel Families
Attach a model family to every agent you build
A model family is a set where different models perform planning, reasoning, and work for sub-agents. Pick a preset in Agent Studio, set model_family in YAML, or override it per run through the API.
Sonnet 5 does the work, Haiku 4.5 handles sub-agents, Opus 5 plans. The same models a run uses when no family is set.
Claude Sonnet 5Agentic coding and autonomous tool use · Jun 20261M in · 128K out · $2 / $10 per MTokClaude Haiku 4.5Fast, low-latency multi-agent orchestration · Oct 2025200K in · 64K out · $1 / $5 per MTok
Gemini 3.7 Flash does the work and sub-agents, Sonnet 5 plans. Fastest and cheapest option that still codes well.
Claude Sonnet 5Agentic coding and autonomous tool use · Jun 20261M in · 128K out · $2 / $10 per MTok
Opus 5 does the work and plans, Sonnet 5 handles sub-agents. Slowest and most capable.
Claude Sonnet 5Agentic coding and autonomous tool use · Jun 20261M in · 128K out · $2 / $10 per MTok
GPT-5.6 Terra does the work, Luna handles sub-agents, Sol plans.
GPT-5.6 Sol throughout. One model for planning, execution, and sub-agents.
Grok 4.6 throughout. Frontier-level agentic scores at mid cost, slower per token than the others.
Kimi K2 Thinking does the work and plans, Qwen3 235B handles sub-agents. Lowest cost per token in the list.
Tooling
Four ways to access Paradime
Whatever you automate, there is a direct way to reach Paradime over HTTP, from the terminal, in Python, or as events over webhooks.
mutation {
triggerBoltRun(slug: "daily-models") {
runId
}
}{ "data": { "triggerBoltRun": { "runId": 15477 } } }GraphQL API
Get full programmatic access to your workspace across Bolt, users, audit logs, and DinoAI agents.
$ pip install paradime-io
$ paradime bolt run "daily-models" --waitTriggered run 15477
dbt build .......... [PASS in 3m 12s]
Run 15477 finished: SUCCESSParadime CLI
Manage Paradime resources and trigger integrations from your terminal, or from Airflow, GitHub Actions, and 20+ other orchestrators.
from paradime import Paradime
paradime = Paradime(api_secret="...")
run_id = paradime.bolt.trigger_run(run_name="daily")>>> run_id
15477Python SDK
Work with Paradime from Python with the official SDK and using every API module.
POST https://hooks.acme.io/paradime{
"id": 1234,
"status": "succeeded",
"commands": [
{ "command": "dbt run", "status": "succeeded" }
]
}Webhooks
Subscribe to bolt.run.started and bolt.run.completed events and forward them to Slack, Jira, Linear, or your own HTTPS endpoints.
Custom Integrations
Add your own custom data products to your lineage
The Custom Integration API extends Paradime's lineage and catalog with external applications. Define your own node types, each with an icon and color. Upload nodes from Python, and connect upstream and downstream dependencies to warehouse tables and dbt™ models. Ready-made examples cover Matillion, Streamlit, and Snowflake Notebooks.
Available with the Paradime Enterprise pack.
Join industry leaders winning with DinoAI agents
Data engineers and analysts at the most AI-native startups and enterprises use Paradime's DinoAI agents everyday to become market leaders.