03 / What we build against
Named interfaces, written out. No logo wall.
A name below means we build against that product's published interface and can read its documentation without a training course. It does not mean a partnership, a reseller agreement, a certification or an endorsement by that vendor, and we hold none of those.
Systems of record
Salesforce, HubSpot, Microsoft Dynamics 365, Zendesk, Freshdesk, Jira, Linear, Monday.com, Asana, Airtable, Notion, Shopify, Stripe, Xero, QuickBooks, Sage, SharePoint, Google Sheets.
Automation and integration
n8n, Make, Zapier, Microsoft Power Automate, Temporal, Celery, cron, webhooks, REST and GraphQL interfaces, SFTP drops, message queues, Postgres, MySQL, SQLite.
Runtime and hosting
Python, Node.js, TypeScript, Docker, AWS Lambda, Google Cloud Run, Azure Functions, Cloudflare Workers, GitHub Actions, and plain scheduled jobs on a server you already pay for.
Model interfaces
OpenAI API, Anthropic API, Google Gemini API, Azure OpenAI Service, Amazon Bedrock, and open weight models run on hardware you control where the data cannot leave.
Where a step can be done by a rule, we use a rule. A language model is expensive, non-deterministic and hard to audit, so it earns its place only on the steps that genuinely need reading and judgement: classifying free text, extracting fields from documents that arrive in twelve different formats, drafting a reply a human then approves.