Solutions · Existing businesses adopting AI
Adopt the harness without replacing your stack.
Keep your ERP, your CRM, your store and your books. The Engine sits beside them as a governed workspace: your agents drive it through an audited action registry, your data enters through connectors and CSV with no model in the path, and every run is priced, ledgered and auditable. Nothing you connect is ever used for training.
01 — What changes on day one
Your agent gets a door. Your data gets a path. Both are audited.
Three workflows, each naming the surface it runs on and what it costs — the product shows the same estimate before every metered call, whether a person or an agent makes it.
01
Mint a token and connect your agent
On Settings → Agent Integration, mint a per-workspace API token: shown once, hashed at rest, rate-limited per token, revocable. Point Claude, ChatGPT or an agent you built at the MCP endpoint and it sees 8 tools — run_tool, lead_search, get_deliverable, search_media, search_events, book_event, radar_summary, report_status. The REST surface is described by a product-true OpenAPI document of 12 paths; only routes that exist are described, and the live action manifest publishes every action’s parameters and credit estimate so an agent can plan and price before it acts.
02
Connect the store, import the rest
Connect your Shopify store from Data Connectors and orders, customers and products arrive on a schedule, mapped into Company Memory by code. Import a lead list from CSV and its headers become research columns; import CRM contacts, leads and deals through the CSV adapter; upload your business numbers as a CSV into the warehouse, where a deterministic Insight Brief flags week-over-week anomalies. Dataset import and export are confirm-class actions — they move whole datasets — and they are audited like everything else.
03
Order the first governed run from the agent
Let the agent call run_tool for a Competitive Landscape Analysis — 800 credits at full depth, cited, in minutes — then poll report_status and fetch the PDF with get_deliverable over a short-lived link. The call meters through the same reserve → settle pipeline as a click in the product, lands as an identical audit row, and is refunded in full if it fails.
02 — Data connectors
Your data, with no model in the path.
The connector ledger below is rendered from the same registry the product reads. Where a provider is not yet live, this page says so, and so does the product.
| Provider | Category | Status | What that means |
|---|---|---|---|
| Shopify | Selling | Sync live | Orders, customers and products fetched on a schedule; no model in the path. |
| QuickBooks Online | Money | Mapper built | Field mapping into Company Memory is shipped; the fetch awaits the provider’s platform registration. |
| Stripe | Money | Mapper built | Field mapping into Company Memory is shipped; the fetch awaits the provider’s platform registration. |
| Plaid | Money | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| PayPal | Money | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| Xero | Money | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| Square | Selling | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| Amazon Seller | Selling | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| WooCommerce | Selling | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| Klaviyo | Marketing | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| Meta Ads | Marketing | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| GA4 | Marketing | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
| HubSpot | Relationships | Registered | Scopes and data-use policy declared; the handshake answers “awaiting platform registration” until the app is registered — never a fake handshake. |
Ingestion is zero-AI by construction
Connector data is fetched, mapped and written into Company Memory by code, not by a model. Card and bank numbers are never fetched. Each provider declares its scopes and a data-use policy, and the CRM-type connectors state the law plainly: stored for your workspace’s own intelligence, fenced by row-level security, never used for training. Connector tokens live in a per-workspace vault under envelope encryption, so one workspace’s connected data can be crypto-shredded on its own.
CSV in, CSV out
Lead lists, CRM contacts, leads, sequences and deals import and export through CSV, with mappings for the CRMs you may already run. Business numbers upload as CSV. The audit trail exports to CSV. The whole Company Memory exports as a checksummed Mind File and imports into another workspace with its derived layers rebuilt. You are never locked in by format.
No AI ever touches the ingestion path, and nothing you connect is ever used for training.
03 — The departments you get
What the workspace does once your agents and your data are in it.
Each department is a real surface in the product, reachable by a person in the product or by an agent through the registry.
Agent Integration
The MCP server with 8 tools, per-workspace tokens, the OpenAPI description and the live action manifest. Three external-agent modes for the AI manager — solo, copilot, replace — so your agent can propose into the queue or hold the tools under the same fences.
Data Connectors and CSV
13 Tier-1 providers registered with scopes and a data policy; Shopify sync live; QuickBooks Online and Stripe mappers built. CSV for leads, CRM entities, business numbers, the audit trail and the Mind File.
Company Memory
14 floors with a 3D Memory Stack, Theory of Mind on counterparties, Brain Dump for your own documents, the Layer Foundry to design a custom floor for 250 credits, and nightly consolidation.
Department of Intelligence
The DILIGENCE ENGINE™ with 15 report types — including data-grounded internal reports at 400 credits that read your connected data and never the web — the Legal Department, Private Investigation and the Executive Daily Briefing.
Daemon Operator
Strategic Objectives with plans drawn as flow charts, a tactical floor with an AI receipt in cents on every job, a five-minute patrol in SQL, a Monday pack, a kill switch and a trust ladder.
B2B DealFlow
Lead Generator by plain-English geography at 8 credits per lead found, the Lead Database, outreach campaigns from your own SMTP mailboxes, RFP Radar with a four-asset proposal kit at 350 credits, and the CRM.
Social & Online Presence
Social Media Poster drafting and scheduling for 12 platforms, Brand Radar across 14 source types with AI-visibility scans, a customer Chatbot trained on Company Memory, and classifieds drafts.
The Admin control plane, for groups
A group with several entities runs them as separate workspaces under one root plane: a tenants table with balance, reserved credits and 30-day burn per entity, a live price sheet, credit gifting, a global broadcast, a cross-entity audit view and Mission Control.
White-label
A division that wants its own chrome takes the partner option: brand name, primary colour, logo, from-email and custom-domain fields, a plan builder and template provisioning. Partner staff see their brand in the hub today; client-facing domains and branded PDFs are roadmap.
04 — Governance
The controls your security review will ask about.
Each of these is enforced in the database or the application layer, not in a prompt, and re-checked in CI.
Row-level isolation
Row-level security on all 167 tables, 183 policies. Claims are the first wall; Postgres is the second. Even the root operator is scoped to the active workspace on tenant tables, and a CI tripwire refuses any policy that re-grants a cross-tenant read.
Audit, append-only
Every action by a person, an agent or an integration lands as one row with its source. Update and delete are revoked at the database grant layer — including from the service role — and blocked again by trigger. Filter by actor, action and date; export to CSV.
Secrets sealed
AI keys and platform tokens are sealed in the application before they touch the database; connector tokens sit under an envelope vault with a per-workspace key. Logs and responses show only the final characters. Secrets are decrypted just in time for one call and never persisted in clear.
In-house AI keys
Every model call runs on New Renaissance Labs’ own keys across 8 providers — Anthropic, OpenAI, Google, xAI, DeepSeek, Qwen, Moonshot, Z.ai — with per-prompt model pins and a capability registry that strips unsupported parameters before dispatch. You manage no provider account; usage is metered to your workspace in credits with a receipt in cents per call.
Gates the server decides
Every action carries a risk class. Safe steps run; confirm-class steps wait for a person. An agent cannot downgrade a class or skip the ceremony by omission, because the server — not the caller — decides whether one was required.
A kill switch per workspace
Pause everything halts the AI manager within one tick. Revoke a token and the agent behind it is out. Factory-reset a workspace and its data is gone while the shell, its seat and its credits remain — with a typed confirmation and no undo.
Identical rows for route, agent and MCP. Same prices, same gates, same audit.
05 — Operating it
Operate it yourself, or have it deployed. Some things are an engagement, not a feature.
Mission Control is the platform’s operations board: the active executor and its heartbeats, manual failover, the cadence of every scheduled task, and the error log, with every change audited. Behind it, 17 scheduled automations drain the queue every minute, patrol every five minutes, consolidate memory nightly and compose the Monday pack. Sign-in today is a magic link plus Google and Facebook; there are no password fields.
The list on the right is what we do not sell as a feature. Each item is built to order on the same substrate — the registry, the ledger, the audit trail, row-level security — and scoped as a project. See Consulting.
| Need | Today | Status |
|---|---|---|
| SSO / MFA / enterprise identity | Magic-link and Google/Facebook sign-in. No SSO, no MFA. | Engagement |
| Live sync beyond Shopify | QuickBooks Online and Stripe mappers built; HubSpot, GA4, Meta Ads and the rest registered. Live fetch is built per provider, to order. | Engagement |
| Private deployment and data residency | One multi-tenant platform under row-level security; a private deployment is scoped as a project. | Engagement |
| Client-facing white-label domains, branded email and PDFs | Partner staff chrome is branded today; clients run on the Engine’s chrome. | Engagement |
| Custom execution workflows | Zero-AI n8n workflows for any back-office process, on the same service tokens the platform’s own workers use. | Engagement |
| Custom report types, prompt libraries and memory layers | The Layer Foundry designs a custom floor in the product; bespoke report types and ontology extensions are built to order. | Engagement |
06 — What it costs
A first month, priced from the sheet.
Growth is $349 a month for 5,000 credits. Below is a plausible first month for a company whose agent runs the research, whose store is connected, and whose legal and memory work has begun. 1 credit is 10¢ CAD.
| Item | Where it runs | Credits each | Quantity | Credits |
|---|---|---|---|---|
| Web-research reports, ordered by your agent over MCP | run_tool · whitepaper.generate · full depth | 800 | 2 | 1,600 |
| Internal reports from connected data | Internal Sales Performance · no web research | 400 | 2 | 800 |
| Prospects researched by geography | Lead Generator · per lead found | 8 | 100 | 800 |
| Legal drafts: DPA, MSA, contractor agreement | Legal Department · jurisdiction required | 250 | 3 | 750 |
| Executive Daily Briefing, one per day | Grounded · never invents a headline | 10 | 30 | 300 |
| Design one custom memory floor | Layer Foundry · one design pass, activation free | 250 | 1 | 250 |
| Month, total | 4,500 | |||
| Growth plan | $349 a month · 1 credit = 10¢ CAD | 5,000 | ||
| Growth covers it with 500 credits left. | 500 |
Plans: Core $149 a month for 1,500 credits · Growth $349 a month for 5,000 credits · Scale $749 a month for 12,000 credits. Packs: 500 for $45 · 2,000 for $159 · 8,000 for $549. A group with several entities funds each workspace from the Admin plane; partners pay the wholesale tier. Detail on the pricing page.
Where next
Other doors into the same harness.
Related
Business owners
What the workspace does once your records are in it: briefing, research, radar, outreach.
Related
Private equity operators
The control plane across many companies, and the roadmap built with design partners.
Related
Partner programme · Consultants
Run every client engagement on a harness — objectives, plans, reports, radar and outreach.
Next step
Point your own agent at a real workspace.
A private demo is a real workspace, provisioned for you with credits and an access password. Mint a token, connect the agent you already use, and read the audit trail it leaves.