Features

Everything the platform does, grouped by the job it belongs to

Grouped by area rather than laid out as one long grid. Open the one you are evaluating; the rest stays out of the way.

Analysis

Where an engagement's material is captured and kept current.

Requirements workspace: The brief every deliverable is derived from, editable in place.
Project and client record: Engagement, client, owner and status, with the deliverable spine attached.
Document-type registry: Twenty-four types, their dependencies and their prompts, administered centrally.
Attention backlog: Derived, not maintained: every deliverable now behind the document it was written from.

The engine reads the engagement before it writes a word of it.

Project classification: What kind of engagement this is, which decides which specialists are worth running.
Discovery model: Requirements, assumptions, open questions, risks and dependencies, extracted and stored as records rather than prose.
Specialist passes: BA, Technical Architect, Security, Estimation and Risk, each scrutinising the ground discovery leaves shallow.
Dependency research: External capabilities the brief relies on, checked rather than assumed.
Quality gate: Every generated document is scored before you see it, and never fabricates detail to raise the score.
Cross-document consistency: The pass that catches the SOW and the estimate disagreeing on feature count.
Paid for once: Analysis is fingerprinted against the requirements and the indexed documents: unchanged sources reuse the model instead of re-running it.

Generation that reads the project rather than a prompt box.

Single-deliverable generation: Write any of the twenty-four types, streamed as it is produced.
Full-pack generation: The core eight in dependency order, as one run you can cancel or retry.
Workspace prompt overlays: House rules layered onto the system prompts, per workspace, without forking a template.
Multiple providers: Anthropic, OpenAI, DeepSeek, z.ai, MiniMax, Groq, OpenRouter — chosen per run.
Run instructions: A focus note per run, and explicit milestone counts and weightings.
Assistant chat: Ask questions of the engagement's own material, with citations.
Web research: Optional external retrieval, indexed into the project's knowledge like any upload.

The client's own material, made retrievable and kept inside its workspace.

Uploads: PDF, DOCX, text and markdown, extracted on arrival.
Chunking and embeddings: Split, embedded and stored as vectors alongside the project.
Grounded retrieval: Relevant passages are retrieved per generation and fed under a budget.
Workspace isolation: Retrieval is scoped to a workspace in the query itself, not by convention.
Reindexing: Re-run extraction and embedding when a source changes.

What happens after the model stops writing.

In-place editing: Every deliverable is editable; a hand edit is a version like any other.
Immutable versioning: Nothing is overwritten. Each generation, edit and restore appends.
Compare: Line-level diff between any version and the one before it.
Restore: An earlier version is copied forward, so restoring is itself undoable.
Reading and paper view: Read on screen, or see the exact page the export will produce — furniture, rules and watermark included.
Diagrams: Architecture and data-model diagrams are drawn on the page, not left as code.

Client-ready files, branded by the workspace rather than the exporter.

DOCX: Word output with the workspace's page furniture, headers, footers and watermark.
PDF: The same document rendered for print.
Markdown: The assembled source, for teams that keep documents in a repository.
Pack export: A ZIP of individually branded files, in reading order, behind a cover page and contents.
Version export: Any historical version, rendered exactly as it was written.
Document format: Page size, margins, fonts, accent colour, logos, watermark strength and confidentiality note, set per workspace.
One layout, three renderers: Word, PDF and the on-screen preview share one set of measurements, so what you approve is what the client opens.

Organisations, workspaces and the people inside them.

Organisations: The billing and identity boundary a customer signs up as.
Workspaces: The data boundary: every project, document and run belongs to exactly one.
Invitations: The only route in for someone outside the organisation.
Google sign-in: Or email and password, with sessions a user can revoke from their own account.
Role-based access: Five organisation roles and four workspace roles, resolved from one capability matrix.
Custom roles: An organisation can define its own, layered on top of the built-in ones.

What a delivery lead needs to see without asking anyone.

Usage analytics: Runs, tokens, cost and duration by provider, model and workspace.
Spend ceilings: A monthly limit that refuses generation rather than billing past it.
Activity ledger: Who did what, to which object, and when.
Audit log: Field names, never values: requirements and client content stay out of it.
Notifications: Runs that finished or failed, and documents that could not be processed.
Subscriptions: Upgrade, downgrade or renew a plan in the product; the period in force and every payment are on the billing page.

Platform-level control for whoever operates the deployment.

Provider configuration: Which engines are available, and which model each defaults to.
Prompt administration: System prompts and format templates per document type, versioned on change.
Model pricing: Rates per million tokens; every recorded cost and ceiling is computed from them.
Document types: Add, disable and reorder the twenty-four types the pipeline produces.
Plans and coupons: Plans, prices, features and discount codes, defined in the product rather than in a config file.
Payment gateways: Stripe and Razorpay, switched on per deployment; secrets stay in the environment and are never returned by an endpoint.

Not on the list

A feature list is more useful when it also says where the edges are.

It writes the documents an engagement needs. Tracking the delivery of that engagement belongs in the tool your team already uses.

Every document type is a business-analysis artefact with its own structure. It is not a blank editor with a model attached.

Nothing is generated, restored or exported without somebody asking for it, and every one of those is recorded.

Create an account, run one engagement through it, and the matrix above stops being abstract.