OpenAI automation, with the agent on your own machine.
Cronable uses OpenAI two ways. A model step sends a prompt to the Chat Completions API and passes the answer to the next job; a Codex job runs OpenAI’s coding agent in a folder on your machine. Either can sit anywhere in a workflow.
A OpenAI workflow on the canvas
Each card is one job. A wire means the next job runs when the one before it succeeds, or when it fails.
Route to Slack
Slack·Idle
After Summarize it (succeeded)
Summarize it
OpenAI·Running
After Support inbox (succeeded)
4s
Support inbox
Inbox·Succeeded
On new email
8s ago
Support inbox
Inbox·Succeeded
On new email
8s ago
Summarize it
OpenAI·Running
After Support inbox (succeeded)
4s
Route to Slack
Slack·Idle
After Summarize it (succeeded)
“When a new email arrives in my support inbox, have OpenAI classify it and summarize it, then post it to the right Slack channel.”
Every OpenAI step Cronable ships
Actions
Steps a workflow can take in OpenAI.Prompt a model
Send a prompt and optional system message to the model you choose, with optional token cap and temperature. The reply becomes the step’s output for the next job.
Get structured output
Give a JSON Schema and the reply must match it, so later steps can read fields directly instead of parsing prose.
Let the model use tools
Switch on only what a step needs: read or write files inside folders you name, fetch pages from hosts you allow, or run jobs you list. With nothing set, the model has no tools.
Use a compatible gateway
Point the base URL at an OpenAI-compatible endpoint, such as a self-hosted vLLM, LiteLLM or Ollama server.
Run Codex in a folder
A Codex job runs the Codex CLI headlessly in a working directory to read and edit code and run commands, in a read-only, workspace-write or full-access sandbox.
Triggers
What starts a OpenAI workflow.OpenAI steps do not trigger workflows; they sit inside them. Start one from a schedule, another job, or any trigger: a new email, a webhook, a new row in a sheet.
OpenAI workflows worth automating first
Inbox triage
The Inbox trigger hands each support email to an OpenAI step that classifies and summarizes it, and Slack routes it to the right channel.
Structured extraction
The text of a receipt or invoice email goes to a model step with a JSON Schema, and the vendor, amount and date come back as fields ready for a sheet or your books. Attachments are not read.
Nightly code chores
A Codex job runs in your repository each night to update dependencies or fix lint errors, and a Slack message reports what changed.
Research with web access
Allow a model step to fetch from a short list of hosts, ask it to check competitors’ pricing pages weekly, and email the summary.
How you connect
Connecting OpenAI
OpenAI connects with an API key you paste; Codex uses the sign-in already on your machine.
- 1Store an OpenAI API key as an OpenAI credential, or set OPENAI_API_KEY in Cronable’s environment.
- 2The key travels only in the Authorization header and is never logged.
- 3For Codex jobs, sign in once with codex login as the user Cronable runs as. There is no key in the job.
Your OpenAI automation, on your own machine
Runs on your own machine
Install Cronable as a Mac or Windows app, or on a server you own. Your OpenAI jobs, secrets and run history stay on that machine by default.
Built from one sentence
Describe the workflow in plain English and the built-in assistant builds the jobs, the schedule and the failure handling for you to review on the canvas. On every plan, including Free.
Unlimited runs
Plans count the jobs you define, not how often they fire. A job that touches OpenAI every hour costs the same as one that runs once a month.
Credentials encrypted locally
Keys and tokens are AES-256-GCM encrypted in a local database, referenced from jobs by name and masked out of logs and run output.
OpenAI automation, answered
Can Cronable run OpenAI models in a workflow?
Yes. An OpenAI step sends a prompt to the Chat Completions API and passes the reply to the next job. It can return JSON matching a schema and, when you allow it, use tools.
Can Cronable run Codex?
Yes. A Codex job runs the Codex CLI headlessly in a folder on your machine, with the sandbox mode you choose, and captures its final message as the run’s output.
Where is my OpenAI API key stored?
In Cronable’s encrypted credential store on your own machine, or in its environment if you prefer. It is sent only in the Authorization header and never logged.
Can I use a local or self-hosted model?
Yes. Set the base URL to an OpenAI-compatible server such as vLLM, LiteLLM or Ollama. There is also a dedicated Ollama job type.
Are AI runs limited?
Cronable does not meter runs, including AI steps. Your OpenAI account’s own usage and billing apply. The limit Cronable sets is on its built-in assistant’s daily messages.
Anything you do twice in OpenAI is cronable.
If you paste the same thing into ChatGPT every week, that prompt is cronable.