Documentation
The operator’s manual.
How the New Renaissance Engine is set up, governed, metered and driven — by people and by agents. Everything here names work the platform ships today. A surface that is built but enabled per workspace is marked Preview; work not yet built is marked Roadmap or left out.
01 — Overview
Three layers, one ledger.
A business harness uses AI to run one company’s operations. A private equity harness runs every company in a portfolio — and the portfolio itself. The Engine is built in three layers, and this manual follows them.
The control plane
Admin and the Partner console: every company an isolated workspace, provisioned in one action, funded, priced from one live sheet, read across the book for balance and burn.
Sections 03 · 13
The company
Per workspace: a Daemon Operator that plans as a flow chart and earns autonomy; departments for intelligence, dealflow, capital and presence; a Company Memory every tool writes into.
Sections 04 · 09
The substrate
One action registry (158 actions) as the single mutation choke point, append-only credit and audit ledgers, row-level security on 167 tables, sealed secrets, and an MCP server for external agents.
Sections 10 · 14
02 — Getting started
Sign in, claim the workspace, name the brand, file an objective.
01
Sign in
Email magic link, Google or Facebook; no passwords exist. A first sign-in provisions your own workspace, with you as owner.
02
Claim the workspace
If an operator created the company for you, open the link, enter the access password and sign in with your own email. The account rebinds to you.
03
Name the brand
One active brand per workspace: identity, channels, voice, knowledge. Channel credentials are sealed on save; only the last four characters show.
Claimable accounts. From Admin → Accounts an operator mints a company in one action: a real login, a workspace on a plan, an owner membership, a credit balance, and a shareable link behind an access password. The operating lead claims it with their own email; revoking the link suspends the workspace. Onboarding survey Preview: when enabled, a new owner’s first visit asks for the business name, website, what you sell and a description, and turns the answers into the brand record and a first note in Company Memory.
The first objective. In Strategic Objectives, ratify the mandate and file an objective with a metric, a target and a deadline. State an intent and the Operator drafts a plan as a flow chart to approve, edit or decline; nothing runs before that. The Setup Guide at /onboarding keeps a checklist per feature; the Agent Setup Kit lets your own AI assistant do the clicking.
03 — Control plane
Admin runs the portfolio. The Partner console runs a nation of clients.
Admin is root-only and invisible to everyone else — a non-root caller gets not-found, never forbidden. Every action here lands in the audit log.
| Admin surface | What it does |
|---|---|
| Tenants | Every company with plan, status, balance, reserved and 30-day burn; CSV export; hard delete behind a typed confirmation. |
| Accounts | Mint claimable companies — login, workspace, membership, credits, link and access password. Revoke suspends. |
| Credits & pricing | Gift credits to any workspace, audited. Set the live price of any action; every tenant’s table updates at once. |
| Broadcast | A global announcement banner to every workspace. |
| Self-heal | The switch for the deterministic health sweep and its incident register, enforced in the database. |
| Prompts, routing, access | Master prompts edited live; per-prompt model pins; a per-workspace matrix of who may see or edit prompts and change models; a ledger of every prompt execution. |
| Audit | The cross-tenant audit viewer: actor, action prefix, date, reference, workspace. |
| Mission Control | Executor status and failover, task cadence, the error log — every action audited. |
Partner console
A partner is a white-label operator: its seats set the brand tokens their staff see, build a plan menu, provision client workspaces from templates with starter credits, and read a monthly wholesale statement at tier pricing. Their clients run on the Engine’s chrome; client-facing theming is Roadmap. See White-label.
04 — Daemon Operator
An AI manager that plans in flow charts and earns its autonomy.
One loop, two working tiers. The reflex tier is pure SQL: every five minutes it checks belief tripwires, error spikes and KPI drift, and most ticks end there at zero tokens. The tactical tier runs only when a reflex tick wakes it: one bounded model call that revises the situation narrative, citing recalled memory; an unsupplied number or an uncited memo is rejected, and memory reaches the model as data, never as instructions. The loop is switched on per company and starts in shadow mode; its own thinking is free today.
Objectives, plans and gates
An objective carries a statement, a north-star metric with target and deadline, key results, a priority and a share of the credit envelope. State an intent and the planner drafts a plan graph bound only to actions in the live registry; anything that contacts the outside world is gated by construction, and every plan ends with a step that tracks results. If the AI planner fails, one of 8 command playbooks produces a reviewable plan instead, free; an AI draft is 15 credits.
Approve, request changes with a note, or decline with a reason the next draft recalls. On approval the steps are recompiled from the graph you saw and each runs unconfirmed, so the live risk class decides: safe steps fire, confirm-class steps park at their gate, and releasing a gate is the only way such a step executes. The canvas allows 12 node kinds from a step library of 20 owner-voice families.
The trust ladder
Autonomy is earned per domain and per reversibility class: R0 easily undone, R1 hard to undo, R2 permanent — and R2 always asks. The promotion test (settled decisions, a Brier score within bounds, zero breaches) is written into the database, and only a decision you consent to raises a level. Pause everything halts the Operator within one tick; Resume restarts it.
| Level | Name | What it may do |
|---|---|---|
| L0 | Watching | Shadow mode. It reads, remembers and proposes nothing. |
| L1 | Suggesting | Plans and proposals land in the decision queue; nothing runs until you say go. |
| L2 | Acting within budget | Safe steps run inside the envelope; money and irreversible steps still wait for a click. |
| L3 | Fully trusted | Expanded autonomy, earned through graded decisions and Brier calibration — demotion is automatic. |
The tactical floor and the cadence
Tactical Operations is the floor — systems board, ops radar, the diary in plain sentences, AI faults, heartbeat, mission clocks, operations and alarm boards — every number query-derived.
| When | What runs |
|---|---|
| Every minute | The job executor drains every company's queue. |
| Every 5 minutes | The daemon patrol checks belief tripwires, error spikes and KPI drift — in SQL, usually for zero tokens. |
| Hourly | Executive Daily Briefing refresh; RFP Radar sweep. |
| Nightly | Memory consolidation: entities, edges, lessons, salience decay, belief re-checks. |
| Monday | The weekly pack: scorecard, decisions taken, reserved queue, risk moves, one question. |
| Monthly · Quarterly | Monthly review and quarterly horizon packs. |
05 — Company Memory
14 floors, one append-only spine, a mind you can carry.
Memory is event-sourced: the spine of events is the truth; entities, edges, lessons and the situation model are derived from it and can be rebuilt. Capture happens at choke points — every queued job, every registry call, the Operator’s own loop — with secrets redacted first. Facts are filed by 17 entity kinds and carry a provenance tier, so a weaker source never overwrites a stronger one. Recall is zero-AI and returns a citation-ready packet.
| Floor | Region |
|---|---|
| 14 | Theory of Mind |
| 13 | Leads & Pipeline |
| 12 | Decisions & Lessons |
| 11 | Customers & Accounts |
| 10 | Cash & Money Operations |
| 9 | Game Theory |
| 8 | Supply Chain |
| 7 | Product & Market Pricing |
| 6 | Industry Trends & Macro |
| 5 | Investment & Valuation |
| 4 | Online Reputation |
| 3 | Competition & Ownership |
| 2 | Risk Analysis |
| 1 | Market Demand |
Intakes
Every research report ends with a machine-readable appendix captured into its floors. Train Now and Increase Intelligence re-run a floor’s feeding report, cost shown first; studying a stored report into memory is 200 credits.
Brain Dump
PreviewDrop files or whole folders. The chain chunks without AI, then summarises, clusters, classifies and integrates — each stage a priced job, estimate shown first. Your documents can correct web-researched beliefs.
Layer Foundry
Describe a new kind of memory in plain English; one design pass (250 credits) compiles an ontology contract — entity kinds, attributes, relations, charts, a 3D scene. Activation is free.
Theory of Mind
For each rival, customer, prospect, supplier, partner and key person: goals with evidence, probability-weighted next moves, our best response, leverage and the test that would falsify the model. Generation is metered.
Consolidation
Nightly, in SQL, at zero tokens: entities and edges derived from the day’s episodes, failures folded into lessons, salience decayed, beliefs re-checked. Memory is archived, never deleted.
Mind File
Export everything the company knows as a lossless, checksummed archive; import it into a fresh workspace and the derived layers rebuild. A per-floor executive summary is 400 credits.
06 — Department of Intelligence
15 report types, a purpose-gated investigator, a legal department.
Pick a report type, name the subject, choose a depth from one to 50 pages, press Generate. The estimate is shown before you order, credits are metered only when the report generates, and a failed run refunds in full. Reports finish in minutes with a live ETA, land as branded citation-first PDFs, and are captured into Company Memory. Price is a pure function of pages: shorter is always cheaper and the bill never exceeds the list price. A deterministic verifier (citations, arithmetic, contradictions, placeholders, structure, dates) can score a finished report, advisory only.
| Report type | What it covers | Credits |
|---|---|---|
| Executive Business Summary | A concise company snapshot with a strategic domain-analysis table. | 800 |
| Competitive Landscape Analysis | Porter's Five Forces, competitor profiles, pricing and talent benchmarking. | 800 |
| Industry Trends Whitepaper | Emerging technology, regulation, consumer shifts and disruptions (PESTLE). | 800 |
| Investment Analysis | Equity-research rigor: variant view, tear sheet, moat, scenarios. | 800 |
| Risk Assessment & Mitigation | Operational, financial, reputational and strategic risks with a scoring matrix. | 800 |
| Human Resources Analysis | Talent, retention, culture, compensation and HR-technology adoption. | 800 |
| Supply Chain Analysis | Suppliers, logistics, manufacturing, distribution and a COGS breakdown. | 800 |
| Market Demand & Customer Analysis | Segments, behaviour, satisfaction, acquisition, retention and LTV. | 800 |
| Public Relations Analysis | Media coverage, sentiment, crisis response and branding review. | 800 |
| Executive Diligence | Career history, legal records, controversies and risk flags for key people. | 800 |
| Game Theory & Strategic Dynamics | Players, strategy sets, payoff matrices, equilibria and ranked moves. | 800 |
| Product & Market Pricing Analysis | The live price map across rivals, margin structure, elasticity and scenarios. | 800 |
| Internal Performance Report | Reads your own warehouse and brand knowledge only. | 400 |
| Internal Sales Performance Report | Requires an active data connection. | 400 |
| Lead Database Intelligence Report | Reads a lead folder; also populates the Leads memory floor. | 400 |
Private Investigation
Citation-first research on a person from the live public web only, purpose-gated: business relationship, fraud prevention and litigation research are allowed; employment, tenancy, consumer credit, insurance and licensing are refused server-side. You pick a jurisdiction mode (US non-FCRA, EU/UK, Canada) and sign a verbatim attestation that is logged; a velocity ceiling holds unusual volumes for review. The report covers disambiguation, profile, sanctions and adverse media, litigation and public records, network, and a red-yellow-green risk panel, with verbatim disclaimers and a public correction channel behind it.
| Order | Credits |
|---|---|
| Due diligence · standard | 1200 |
| Due diligence · deep | 2800 |
| Background check | 600 |
Legal Department
26 document types on 6 shelves, 250 credits a draft; the governing jurisdiction is mandatory. Drafts are written from Company Memory and the brand record, leave bracketed blanks rather than invent a fact, and end with points to confirm with counsel. A non-removable disclaimer sits in the body, footer and screen; nothing is signed, filed or sent.
| Shelf | Documents |
|---|---|
| Contracts & Commercial | Mutual NDA · One-Way NDA · Master Services Agreement · Statement of Work · Consulting / Professional Services Agreement · Supplier / Vendor Agreement · Reseller / Referral / Partnership Agreement · Software / IP Licence Agreement |
| People & IP | Employment Agreement / Offer Letter · Independent Contractor Agreement · IP Assignment Agreement · Advisor Agreement |
| Policies & Compliance | Terms of Service · Privacy Policy · Data Processing Agreement · Acceptable Use Policy |
| Corporate & Capital | Letter of Intent · Investment Term Sheet · SAFE / Convertible Instrument · Shareholder / Founders Agreement · Board / Corporate Resolution |
| Disputes & Notices | Demand Letter · Cease & Desist · Contract Termination Notice · Settlement & Release Agreement |
| Custom | Describe what you need |
Executive Daily Briefing
A daily scan in three categories — local trends, industry trends, competitive moves — from your location and interests; 10 credits per generation. Every item links a live source; zero items is an honest outcome.
Event Finder
A discovery profile — city, radius, interests, exclusions, price ceiling — and a grounded discovery run, daily or on demand, returning real, dated, fit-scored events. Mark each saved, booked, attended or dismissed.
07 — B2B DealFlow
Find buyers, reach them from your own mailbox, track the deal.
Lead Generator and Lead Database
Describe who you sell to and where. Auto-fill pre-fills the brief from Company Memory; a geography pre-flight turns a plain-English area into a city list with a credit estimate before anything is spent. The run designs the researchable fields, lists real businesses in each city and researches each one with citations. A lead costs 8 credits, charged only for leads found. A field that cannot be found stays blank — no invented emails, phones or websites — and email status stays unverified until a human marks it.
Leads score from zero to one hundred with a rationale; qualified leads convert into the CRM on their own. Import a CSV, merge lists and back-fill missing cells, save field specs as templates, draft a first email per lead for your approval. Nothing sends from here.
Outreach campaigns
Connect your own mailboxes over SMTP; credentials are verified live, then sealed. Each campaign writes a fresh plain-text email per lead from your prompt and that lead’s research, drips it inside your window with jitter, rotates across your sender pool and ramps each mailbox from a low daily cap. Preview, then Activate — the engine never activates anything itself. Every message carries a sender footer and a one-click unsubscribe; bounces and unsubscribes are never re-targeted. Campaigns are single-touch; the Sequences builder designs multi-step cadences but does not send them.
RFP Radar and Proposal Studio
PreviewSaved natural-language searches swept on your schedule; 14 official procurement feed providers registered. Per opportunity: a four-asset proposal kit — one-pager, deck, sales cheat sheet, full proposal — for 350 credits, a standalone cheat sheet for 150, department packs, downloads in the buyer’s colours. Upload past proposals and the studio learns your house style as machine-checked rules. Nothing sends.
CRM
Contacts de-duplicated on email with a fit score, rationale, tags and source; deals by stage — New, Contacted, Qualified, Won, Lost — with totals per column; a timeline per contact. CSV import and export.
Preview on RFP Radar refers to official feed ingestion; grounded sweeps, the inbox and the proposal kit run today.
08 — Fundraising & Investments
Raise capital, or hunt companies to acquire.
The capital wing is enabled per workspace Preview. Every number in it is deterministic — the fundability score, the valuation engine, the heat score, the thesis ranking — and a model is only ever asked to narrate a result or gather evidence.
Capital Command and Investor Finder
A fundability score from zero to one hundred built from financial-intake completeness, memory coverage, report freshness, valuation scenarios and room assets; every unearned point is a named gap with the link that closes it. Investor candidates come from live web research with citations — never a licensed database — and fit is scored on stage, sector, geography, cheque size, thesis overlap, portfolio adjacency and recency, each with evidence and a source URL. Prepare outreach hands a draft to the outreach desk; nothing sends.
Valuation Workbench
Scenarios, drivers and live outputs recomputed in the browser on every keystroke by a pure engine; the server stamps identical numbers on save. A metered narrative only describes the figures. Default currency is CAD; open it with a deal id to price a buy-side target.
| Method | What is computed |
|---|---|
| Discounted cash flow | Per-year FCF build; discount rate by build-up or WACC when debt is present; terminal value by Gordon growth, exit multiple or a blend. |
| Multiples | Revenue, EBITDA or seller’s discretionary earnings × multiple, with weights and comparable notes. |
| Unit economics | LTV, CAC, LTV/CAC, payback months, contribution margin, breakeven units — null when an input is missing, never guessed. |
| Round modeller | Pre- and post-money, new investor share, option pool pre or post, SAFE conversion at the better of cap and discount. |
| Sensitivity, football field, blend | A discount-rate × terminal-growth equity matrix; per-method low/mid/high ranges; a weighted blend. |
| Assumptions ledger | Every driver carries its provenance: owner, memory, default or preset. Presets pre-fill and never lock. |
Data Rooms
Mint a room with the cost shown first; fresh reports are reused free. One tokened link, optional password and expiry, revoke at will. Each visitor sees a text watermark with their email; views, dwell, return visits, downloads and questions roll into a heat band.
Investment Thesis
Mandate, sectors, geographies, size bands, deal types, must-haves, red lines, synergy hypotheses, hold horizon, target IRR and scoring weights. Nothing on the buy side runs until a thesis is saved.
Wolf Hunt board
PreviewScent → Stalk → Ambush → Feast, then Eaten or Passed. Cards carry fit signals with sources and a link to price the target; the database refuses an illegal stage move. The one-click hunt — a grounded origination sweep screened against your rulebook — and the structure sandbox with its deal memo run from the investor engine.
09 — Social & Online Presence
Draft and schedule everywhere. Publish where a connection exists. Listen to all of it.
Social Media Poster
Write the generation prompt once; set a real schedule — weekdays, clock times, timezone — that stamps the publish time on every post; prove the prompt with Generate now, then arm it. The composer fans one post out to 12 platforms (Facebook, Instagram, TikTok, Bluesky, Mastodon, X, LinkedIn, Pinterest, Tumblr, YouTube, Reddit, Threads) at each platform’s own length, with a queue, a media library and A/B arms. A publish gate checks banned terms, required disclaimers and hashtag caps before anything can leave.
Publish without asking me is off by default; every generated post lands as a draft a human marks ready. Publishing itself runs through Facebook and an in-app Reddit publisher Preview; for the other platforms the queue holds the post and nothing leaves.
Brand Radar
Watch terms — brand, hashtag, competitor, AI-visibility — across 14 source types plus any link you paste. A sweep is one grounded web-search pass; walled places are searched with scoped queries, never read directly. Judge.me reviews are read in full with the reviewer’s own star rating as sentiment; Reddit is queried by API when you connect a credential. Every source shows when it was last swept and what it found.
The dashboard computes a reputation score, spike detection, share of voice, platform and location breakdowns and a deterministic memory-impact view. AI-visibility scans ask every connected provider your saved questions and record who it names. Drafted replies never auto-send.
Chatbot
PreviewTrain once (150 credits) to compile permitted memory into a customer-safe brief you read before going live; each visitor turn is 2 credits. Public, unlisted, password or domain-locked; blocked topics refused before the model; spend ceilings; share link or embed. No web search, no invented answers.
Classifieds
Batch-draft listings for Kijiji, Craigslist and Facebook Marketplace. These marketplaces forbid automated posting, so you get each platform’s posting link, a copy button and a mark-posted step.
Ads experiment design
Draft experiments — objective, budget, duration — and creative variants per platform, with AI creative design as a metered step. Launching to an ad account is not part of the product.
Personalized Email Campaign
PreviewDescribe a campaign in a sentence; a planner builds a dated twelve-month calendar from a real holiday table and writes each email into a house template. Consent-gated audiences, a verified sending domain, an arming ceremony, paced sending, a bounce-and-complaint breaker.
10 — Credits
The price before the run. Reserve, settle, release.
One credit is 10¢ CAD. Chat with the Operator, compiling a plan and the Operator’s own thinking are free; runs that produce work are priced, and every metered action shows its estimate before you confirm. The estimate is reserved; when the run finishes the actual units are settled, clamped at the reserve, and the unspent part returns; if it fails the reserve is released in full. An overrun is absorbed by the platform. Every AI call leaves a receipt in cents. Purchases are granted by a signature-verified webhook and reconciled nightly; a cancelled plan keeps the credits already granted.
| Plan | Per month | Credits |
|---|---|---|
| Core | $149 | 1,500 |
| Growth | $349 | 5,000 |
| Scale | $749 | 12,000 |
| Pack | Price |
|---|---|
| 500 credits | $45 |
| 2,000 credits | $159 |
| 8,000 credits | $549 |
| From the price sheet | Credits |
|---|---|
| Research report, full depth | 800 |
| Internal, data-grounded report | 400 |
| Legal document | 250 |
| Due diligence · standard / deep | 1200 / 2800 |
| Background check | 600 |
| Lead dossier, per lead found | 8 |
| RFP proposal kit / sales cheat sheet | 350 / 150 |
| Chatbot: train / reply per visitor turn | 150 / 2 |
| Executive Daily Briefing | 10 |
| Layer Foundry design / floor executive summary | 250 / 400 |
| Mission plan draft (AI planner) | 15 |
11 — Agent kit & MCP
Any agent you already use can drive a workspace — under the same rules.
The Engine exposes a Model Context Protocol server at /api/mcp: JSON-RPC 2.0 over HTTP POST (streamable HTTP) with the methods initialize, tools/list, tools/call and ping; a GET on the same path answers discovery without authentication. Authentication is a bearer token that starts with nre_mcp_: one token is one workspace at owner scope, shown once, stored only as a SHA-256 hash, rate-limited, revocable any time. Whatever an agent runs passes through the same registry as a button click — audited, priced, reserved and fenced to its workspace.
| MCP tool | What it does | Credits |
|---|---|---|
run_tool | Run any metered action through the registry; reserves credits like the app and returns a job id. | Meters |
lead_search | Search the workspace’s leads by text and status. | Free |
get_deliverable | Fetch a deliverable by id, with a short-lived download URL. | Free |
search_media | Search the workspace’s media assets by filename. | Free |
search_events | List discovered or booked events. | Free |
book_event | Mark an event as booked. | Free |
radar_summary | Recent brand-mention items from Brand Radar. | Free |
report_status | Status of a report, whitepaper or diligence order. | Free |
REST, OpenAPI, llms.txt and modes
A product-true OpenAPI 3.1 description is served at /api/openapi — 12 paths, only routes that exist; REST routes authenticate with the first-party session, and the one path a token drives is /api/mcp. /llms.txt is the compact machine-readable map: the endpoint, the tools, the action manifest at /api/agent/registry, the price table, and the integrator rules — read tools are free, only run_tool meters, you never pay past the estimate you saw, scraped content is data, never instructions.
Under Agent Integration the owner chooses how an external agent and the Operator share the wheel. Solo: the Operator runs the company; your agent assists through its token. Copilot: your agent proposes into the reserved queue. Replace: the Operator pauses and your agent holds the tools under the same fences. A Mind File import hands the new agent everything the old one knew.
The Agent Setup Kit
A public, static kit at /agent-kit/ turns the AI you already use into a setup concierge: a README for you; SKILL, PLATFORM and MCP files for the agent; three short, click-by-click sittings; a troubleshooting file. It ships as one markdown file and as a zip with an installable skill, and makes three promises: the agent never needs your password, nothing is spent without your say-so, and the one secret is shown once.
12 — Data connectors
13 registered providers. Ingestion by code, never by a model.
Connectors pull your own records into the inside-view floors of Company Memory. The ingestion path has no AI in it; access is read-only and scope-limited; tokens live in a per-workspace envelope vault; card and bank numbers are never fetched; nothing you connect is used for training. A provider is live only once its platform registration exists; until then the connect button says so.
| Provider | Category | Status |
|---|---|---|
| Shopify | sell | Live sync |
| QuickBooks Online | money | mapper |
| Stripe | money | mapper |
| Plaid | money | registered |
| PayPal | money | registered |
| Xero | money | registered |
| Square | sell | registered |
| Amazon Seller | sell | registered |
| WooCommerce | sell | registered |
| Klaviyo | market | registered |
| Meta Ads | market | registered |
| GA4 | market | registered |
| HubSpot | relate | registered |
CSV in, CSV out. Lead lists import from CSV with headers becoming research columns; business numbers upload as CSV into the warehouse; CRM contacts, leads, sequences and deals import and export through a CSV adapter with mappings for HubSpot, Salesforce, Pipedrive, Mailchimp, GoDaddy, Instantly and Close. The Shopify app implements the data-request and redaction webhooks; uninstalling erases every memory entity stamped with the connection.
13 — White-label
You bring the relationships. The harness runs the clients.
A partner record carries a client-facing brand name, logo, primary colour, from-email, a custom-domain field and a wholesale tier. Partner seats build a plan menu, provision client workspaces from engagement templates — generic SMB, e-commerce, local service — with starter credits, and read a monthly statement of wholesale credits owed at tier pricing. A brand-leak scan in CI fails the build if the in-house name is hard-coded on a tenant-facing surface.
Honesty note: today the partner’s clients run on the Engine’s chrome and PDF theme; client-facing domain theming, branded email and branded PDFs are Roadmap. The programme and its economics are at /white-label.
| Wholesale tier | Partner pays |
|---|---|
| Tier 1 | 50% of retail |
| Tier 2 | 45% of retail |
| Tier 3 | 40% of retail |
14 — Security
Governance in the database, not in the prompt.
The full statement — claims and where each is enforced, tenant isolation, the money path, auditability, compliance features, the 8 AI providers, platform security and what we do not claim — is on the Trust & Security page. In short: every mutation passes one registry; the audit log and credit ledger are append-only at the grant layer; row-level security covers all 167 tables with the root bypass removed; no language model touches money; secrets are sealed; sign-in is magic link, Google or Facebook.
15 — Glossary
The words this manual uses.
| Term | Meaning |
|---|---|
| Operator | The person or firm running the portfolio — the platform root or a partner seat. Operators provision, fund, price and audit companies. |
| Workspace | The tenant: one company, one isolated set of rows, one ledger, one audit trail, one AI manager. Every claim and policy is keyed to it. |
| Brand | The company the workspace runs — name, voice, knowledge and channel credentials. One active brand per workspace. |
| Mandate | The AI manager’s rules of engagement: mission, brand voice, never-do list, reserved matters, the owner’s weekly time budget. Only an owner ratifies a version. |
| Objective | A strategic goal with a north-star metric, target, deadline, key results and priority. Plans serve objectives. |
| Plan | A flow chart of steps drafted from an intent and approved before anything runs; steps are recompiled from the graph you approved. |
| Gate | A step that waits for an owner’s click because its action is confirm-class. Approving a plan never approves a gate. |
| Pack | The weekly briefing: scorecard, decisions taken, the reserved queue, risk moves and at most one question. Silence applies the mandate’s default. |
| Floor | One of the 14 platform regions of Company Memory — Market Demand at the foundation, Theory of Mind at the crown. |
| Layer | A custom memory region you design in the Layer Foundry. It sits beside the platform floors and travels in the Mind File. |
| Credit | The unit of metered work: 10¢ CAD at retail. Chat and the manager’s own thinking are free; runs that produce work are priced. |
| Reserve | The estimate taken off the balance when a job is queued — the most a run can ever cost you. |
| Settle | The actual charge when a run finishes, clamped at the reserve. Unspent reserve returns to the balance. |
| Release | The full reserve returned when a run fails. One job, one reserve, one settle or release. |
Next step
Read the manual in a live workspace.
A private demo is a real workspace, provisioned with credits and an access password. Bring the questions this page raised.