Tools & resources reference

Add a contact to the CRM.

Add a company to the CRM.

Create a deal in the pipeline.

Create a task/follow-up, optionally linked to a contact.

Create a customer project (optionally tied to a company/contact).

Set the tech stack for a project (replaces all existing items).

Add a repository to a project.

Set environments for a project (replaces all existing).

Connect a project to an entity in another module, building the cross-module graph.

Add a calendar event (auto-categorized).

Create a new design system.

Add a design token to a design system.

Add a brand/usage rule to a design system.

Update an EXISTING design token (by its id) on a design system — change its value, name, category or description. Use this to retune a brand (e.g. change the accent colour) instead of adding a duplicate token. Find token ids via the resource…

Create a new React component in the org's shared component library and return its id. YOU write the component yourself — the system does NOT generate it for you. Write self-contained, TOKEN-AWARE TSX: use var(--ds-accent), var(--ds-surface),…

Update an EXISTING shared React component (by its id). Only the fields you pass are changed. When changing code, keep it TOKEN-AWARE (var(--ds-*)) so the component stays re-themeable — see sstm://component/recipe. Find component ids via sstm://components.…

Validate AI-generated code against reality BEFORE returning it to the user. Grounds every import against the real npm/PyPI registry (catches hallucinated packages), runs a syntax/bracket heuristic, and — when projectFiles is supplied — verifies that…

Record YOUR review findings on a project's REAL output (components, wireframes, design system, documents, the project as a whole). This is how the connected AI's review reaches the human: YOU read the project (start with the resource…

Mark a single project review finding as resolved (status → resolved). Use after a finding has been addressed so it stops showing as open in the human's Health view. Org-scoped.

Create a new, empty wireframe in the org's wireframe library and return its id. The wireframe starts with ONE empty screen sized for the chosen platform (web → desktop, app → mobile). This is step 1 of building a wireframe. After creating, READ the…

THE MAIN WIREFRAME WRITE TOOL. Replace a wireframe's entire screen structure with one you have designed yourself. Before calling, READ sstm://wireframe/recipe (the data model + full vocabulary + a worked example) and, to extend an existing wireframe, READ…

Attach (or detach) an org design system to/from an entire wireframe. The design system applies to the whole wireframe, not to individual screens or sections. Pass design_system_id to attach, or null to detach.

Post a message INTO a team chat channel as the connected AI, so you become a participant in the conversation (humans see it live in the chat rail and the /chat page). Use this to answer a question, share a finding, or continue a thread you have read. To…

Run an infrastructure operation on the org's CONNECTED Cloudflare account (via their BYOC OAuth connection). Describe the goal in natural language (e.g. 'list all zones and their plan', 'create an A record api.example.com → 1.2.3.4', 'purge the cache for…

List of the organization's design systems.

All design tokens (with design_system_id) across the org's systems.

Rules the AI must follow (e.g. content restrictions).

Fonts the org has installed/uploaded for use in brands.

Color palettes the org has created for use in brands.

The organization's contacts.

The organization's companies.

Deals in the pipeline with stages and value.

Customer projects with their company/contact and status.

The org's wireframe projects (name, platform, screen count, linked project).

THE RECIPE for building wireframes in this system: the full data model (how screens → sections → cells → blocks and links nest), every vocabulary (the 6 layouts with cell counts, the 4 heights, the device ids, all 75 blocks grouped by category with…

ONE wireframe's full structured document (parsed screens + sections + cells + links + directives, plus meta and the attached design system id). Read this to see the current state before extending it with sstm_wireframe_build. Replace {id} with a wireframe…

The cross-module wiring: which entities (companies, deals, events, design systems) each project is connected to.

Everything an AI needs to build for ONE project, bundled: the project, its active design system (tokens + rules + resolved --ds-* brand CSS), linked components / interactions / layouts, and its wireframes (summarized). Replace {id} with a project id from…

THE REVIEW RECIPE for ONE project plus its current state: HOW to review in sSystm (read sstm://project/{id}/context and the linked artifacts, judge them against the design system through the lenses, then record findings with sstm_project_review_record),…

The organization's documents (the writing source of truth). Each entry includes id, title, project_id, updated_at and an extracted structure summary (outline, decisions, requirements, tasks, questions, todos, mentions, wordCount). The full HTML body is NOT…

One document: metadata (id, title, icon, template_id, project_id, created_by, timestamps), the full HTML body, and its extracted structure (outline, todos, decisions, requirements, tasks, questions, notes, mentions, wordCount). Replace {id} with a document…

Tasks and follow-ups with due dates.

Calendar events with auto-category (meeting/call/deadline …).

The org's shared React component library (id, name, category, description, tags, code length, timestamps). The full component code is NOT included here — read sstm://component/{id} for one component's code.

One shared React component: metadata (id, name, category, description, tags, timestamps) and its full source code. Replace {id} with a component id from sstm://components.

THE RECIPE for building a great sSystm React component: the full --ds-* token vocabulary (what re-themes for free), the categories, the prop convention, the rules for writing token-aware self-contained TSX, and a complete worked example. Read this BEFORE…

THE RECIPE for building a design system via MCP: the token vocabulary, the exact token NAMES that drive the brand engine (--brand-*/--ds-*) — accent, bg, surface, border, text, muted, radius, font and friends — how those tokens then automatically theme…

The organization's routines — automations and reminders (org-level). Each entry includes id, title, enabled, schedule (kind + time/everyMin/days), channels, nextFire and note. Use this to understand what recurring reminders and automations the org has…

The organization's invoices (billing documents of kind=invoice). Each entry includes id, number, status (draft|sent|paid|overdue|…), title, company_id, project_id, issue_date, due_date, currency, tax_rate and the computed subtotal/tax/total. Use this to…

The organization's quotes (billing documents of kind=quote). Each entry includes id, number, status (draft|sent|accepted|declined|…), title, company_id, project_id, issue_date, valid_until, currency, tax_rate and the computed subtotal/tax/total. Use this…

The organization's service/product catalog (reusable billing line items). Each entry includes id, kind (service|product), name, description, unit (e.g. hour|piece|fixed) and unit_price. Use this to understand the org's standard offerings and their pricing…

The recent messages in ONE team chat channel, so you can READ the conversation before replying. Each entry is { id, userId, name, text, ts }. Channels are free-form: 'general', 'random', 'dev', or 'project-<projectId>' for a project-scoped conversation.…