Barndoor Automations cloud
Publish a Blueprint and it's callable as an API or an MCP tool, with access controls and run history.
API · MCP server · scheduleAI workflow builder · powered by open-source Frags
The model thinks.
The runtime decides.
Frags runs your workflow like code: deterministic steps where you need certainty, fenced-in model sessions where you need judgment, and a typed contract on each output, by default. Write it in minutes. Run it anywhere.
frags run customer-analytics.fml --llm clauderequire mcp BigQuery
session("server_breakdown") {
use mcp BigQuery
+ Run this BigQuery SQL to get per-server
stats for {{ .params.customer_name }}
schema $ServerStats[]
}
session("daily_rollup", after="server_breakdown") {
use mcp BigQuery
schema $DailyRollup[]
}
# top_servers · tool_breakdown ·
# user_activity · error_patterns
session("executive_summary", after="error_patterns") {
- Synthesize an executive health report.
schema {
overview: string
risk_flags: string[]
action_items: string[]
}
}Calls the tools you already run
- Slack
- Google Calendar
- BigQuery
- Jira
- Gmail
- Google Drive
- Google Analytics
- Postgres
…and any MCP server or OpenAPI spec.
Determinism, declared
You control how much your Blueprint uses the model.
Every step in a Blueprint says exactly what the model may see, touch, and return — and which steps it never touches at all.
Code does the mechanical work.
Tool calls and functions run as plain code, before or between model sessions. Fetching records, computing dates, rendering an email — none of it is left to a prompt.
Same input, same call, every run.
session("missing-contacts", after="calendar-details") { call("queryContact") { emails = $(context.calendar_details.external_emails) }}Authoring
Written, not wired.
Drag-and-drop builders hide the logic in a canvas of settings panels. A Blueprint puts the whole workflow in one file you can read top to bottom.
- Start from a sentence
Describe the job in plain English. The assistant explores your connected tools and writes the FML.
- Review it like code
A Blueprint is one readable file. Diff it in a pull request instead of clicking through panels.
- Version it with your service
It lives in your repo, next to the code that calls it, with the history to match.
- Check it before it runs
Compiled and schema-checked up front, so a broken step is caught before it reaches production.
AI-native backend services in under 10 minutes.
Describe the workflow in plain English. Barndoor Automations sub-agent explores your connected tools, writes the FML, fixes its own routing, and hands you a deterministic, schema-validated service — callable over API and MCP. No boilerplate. No glue code.
Sub-agent mode
Bring your own LLM key
Web IDE + VS Code
API + MCP callable
Consumers
Interfaces & triggers
Consumers
Interfaces
Triggers
MCP
API
Blueprint
AI automation, defined as code
Inference exactly where the process needs judgment—deterministic everywhere else.
Declare its inputs, decisions, tools, sessions, and output schema in one typed, executable blueprint.
Distribute
API · MCP server · Schedule
Govern
On Barndoor: policy on every call
Schema-validated output
Kept in run history
Runtime connections
Tools, data & models
MCP servers
Databases & APIs
LLMs
On Barndoor: the AI gateway
Your identity provider’s roles, and policy on every model and tool call
SSO / SAML
DLP from the gateway
Not templates. Executable Blueprints.
See what actually runs.
Real workflows, drawn from the Blueprints that run them. Every step, scope, and schema below is in the file.
Customer Health Report
Six live queries become one account-health report grounded in product usage, activity, and errors.
Not just SaaS
Runs wherever your data lives.
The runtime is open source. Use our cloud when it's convenient and your own infrastructure when it has to be — the Blueprint doesn't change.
Your infrastructure
Install the open-source Frags CLI and run Blueprints in your VPC or on premises. No hosted runtime dependency.
brew install fragshq/tools/fragsInside your service
Embed the Frags Go library and run Blueprints in-process, behind your own API.
Go library · in-processBring your own LLM
Pick the model per run — Claude, GPT, Gemini, or a local model through Ollama — with your own keys.
frags run brief.fml --llm claudeThe architectural difference
Automation moves work. A Blueprint makes it dependable.
They work together, too: any existing app, agent, or automation can call a Blueprint.
Canvas workflow toolsBlueprint
What is the model doing?An AI step inside a chain of app actions.Scoped sessions with the context you pass, the tools you enable, and an output schema.
How is it built?Built on a canvas, by hand or from a prompt.One file, written from plain English, reviewed like code and checked before it runs.
What does the caller get?Whatever the last step returns.JSON in the shape the Blueprint declares, over API or MCP.
Where can it run?Wherever that platform runs: its cloud, or a self-hosted install of the whole platform.Barndoor Automations cloud, your VPC, on premises, or embedded in your service.
Common questions
Built to fit your stack.
What makes a Blueprint AI-native?
Model sessions are part of the workflow definition. Each session declares its own dependencies, the tools it may call, its scoped context, and the schema its output must match. Frags orchestrates them alongside deterministic calls.
Can my existing tools call a Blueprint?
Yes. Barndoor Automations exposes a Blueprint through an API or MCP, so apps, agents, chat clients, and other workflow tools can use its result.
Can I run it on premises?
Yes. Frags is open source, with a CLI and a Go library. Run a Blueprint on your own infrastructure without a hosted runtime dependency.
Which models can I use?
Frags supports model choice at run time, including Claude, GPT, Gemini, and local Ollama models. You bring your own keys for hosted models.
Your next workflow can be infrastructure.
Write the Blueprint, keep the model inside a governed runtime, and give the result a real interface.