Choose the workflow
Start with the work, not the model.We identify one practical workflow: documents, support, intake, internal knowledge, code review, or another repeatable process where local AI could save time or reduce exposure.
How it works
A local AI appliance your team can understand.
Foundry sits beside the tools you already use. It runs the AI work that touches sensitive data on your own Apple Silicon hardware, then hands outputs to your team for review.
1choose workflow
2install node
3connect tools
4local first pass
5human approves
6dashboard visible
The simple version.
Put the AI node in your office
Foundry runs on Apple Silicon.A Mac Studio sits on your premises and connects to your network like normal office hardware. It does not need a server room.
Install and configure Foundry
We set up the stack.Foundry is configured with the right model profile, local runtime, workflow harness, capacity guardrails, and dashboard. For most buyers, the harness is either OpenClaw for workflow orchestration or Hermes for messaging/integration.
Connect the existing tools
Your systems keep doing their jobs.Email, helpdesk, GitHub, shared drives, case management exports, and document folders stay in place. Foundry reads selected inputs and sends outputs to a review queue or agreed destination.
Foundry drafts the first pass
Read, classify, extract, search, draft, or review.The AI processing happens locally on the Mac Studio. For sensitive workflows, the content does not need to go to OpenAI, Anthropic, Google, or another cloud AI API.
A human approves
No important action is automatic by default.Foundry queues drafts, flags, summaries, extracted data, or review comments. A person checks the work before it is sent, accepted, filed, or acted on.
The dashboard shows the state
You can see whether it is healthy.The dashboard shows runtime health, selected model profile, capacity status, workflow status, queues, and telemetry posture.
Your tools stay familiar. The AI boundary changes.
Existing work enters
- Email attachments
- Shared folders
- Helpdesk tickets
- Case files
- Internal documents
- Pull requests
Foundry local AI node
- Classifies
- Extracts
- Searches
- Drafts
- Reviews
- Flags exceptions
- Logs status
Human review
- Confirm extracted data
- Approve draft replies
- Check matter gaps
- Read cited sources
- Review PR comments
- Decide next action
AI processing stays on your local hardware for configured workflows.
Foundry does not replace your business systems.
Moves local
- AI processing on selected sensitive documents
- Ticket reading and response drafting
- Internal knowledge search over private documents
- First-pass code review
- Workflow triage and exception flagging
Usually stays where it is
- Email delivery
- Case management / CRM / accounting systems
- Helpdesk platforms
- GitHub/GitLab/Bitbucket hosting
- Cloud backups
- Video calls and collaboration tools
- External research
The AI that touches sensitive work moves local. Everything else stays where it already works.
From fit check to first workflow.
| Phase | What happens | Typical time |
|---|---|---|
| Fit Review | We review your workflow, data sensitivity, volume, tools, and expected economics. | 30–60 minutes |
| Recommendation | You get a practical yes/no and the best first workflow to test. | After review |
| Setup | Foundry is installed and configured on suitable Apple Silicon hardware. | 1–2 days for supported setups |
| Testing | Sample documents, tickets, queries, or PRs are run through the workflow. | Included in setup |
| Handover | Your team learns the review queue, dashboard, and support route. | 1–2 hours |
| Live | Foundry runs the agreed workflow with human approval gates. | Ongoing |
Timelines depend on access, hardware readiness, workflow complexity, and data quality.
Designed around control, not blind automation.
- The configured AI processing runs locally on your hardware.
- No cloud AI API is required for the workloads moved to Foundry.
- Outbound telemetry can be disabled and shown in the dashboard.
- Original documents can be preserved with hashes and extraction logs.
- Human approval is required before customer-facing, legal, financial, or operational actions.
- Your team keeps the existing systems of record.
- Foundry is scoped to agreed workflows, not open-ended autonomous action.