Documentation

Understand how Morrow works.

How to set up your company, run a decision cycle, bring in existing work, manage evidence, choose a model mode, and understand billing, security, and support.

Overview

A concise, restrained reference.

Everything below reflects Morrow's current production behavior. Where a capability is limited or deferred, we say so plainly. For a signed-in walkthrough, open the workspace and start a first decision cycle.

Getting started

Create an account, attach a company, and begin your first decision cycle from a natural-language problem.

Create an account: Sign in with Google, or sign in with email and confirm the address from the message Morrow sends. Passwords are checked against a leaked-password database on signup and change.
Create or attach a company: Each company is an isolated tenant with its own objective, evidence, cycles, decisions, and members. You can create a new company or attach an existing one from the Companies dialog.
Begin with a problem: Describe a commercial problem in the composer in plain language. Morrow decides whether to ask focused questions, reuse existing context, or move directly into diagnosis.
Bring in existing context: Use Connect existing project to import work from ChatGPT, Claude, Gemini, Notion, or uploaded files. Imported material becomes candidate context, not automatically confirmed evidence.

How Morrow works

A chat-first surface backed by a structured, versioned decision system. The founder retains approval authority at every step.

The four phases: Understand, Diagnose, Test, Decide. The chat presents each phase in plain language; the underlying records — objective, evidence, constraint, experiment, brief, outcome, decision — are versioned and preserved.
Canonical records: Objectives, evidence, constraints, experiments, implementation briefs, outcomes, and decisions are the authoritative history. The conversation is a projection over these records, not a replacement for them.
Founder authority: Morrow proposes; the founder approves or modifies. Objectives, constraints, experiments, and final decisions are only ever recorded after explicit founder approval.
Weekly Operating Review: A periodic surface over active cycles that surfaces stalled work, unreviewed evidence, and decisions ready to close.

First decision pass

Your first cycle establishes the objective, surfaces evidence, names the constraint, and designs a small measurable experiment.

Focused questions: Morrow asks only the questions needed to reach a defensible objective. Existing context is reused where possible.
Evidence, provenance, and contradictions: Every piece of evidence carries provenance. Supporting and contradicting evidence are shown side by side. Missing information is called out rather than filled in.
Strongest alternative: Alongside its recommended constraint, Morrow surfaces the strongest alternative interpretation so you can evaluate both.
Smallest credible experiment: Recommended experiments are the smallest that could credibly move the constraint. Success and failure thresholds are frozen at approval so results cannot drift.

Companies and collaboration

Companies are the tenant boundary. Members hold explicit roles; owners and admins can archive; historical records are preserved.

Switching companies: Switch the active company from the Companies dialog. Chat, workspace, and history are always scoped to the active company.
Roles and invitations: Each member holds a role on the company. Invitations are sent by email and expire; membership status is checked on every server request.
Archive and restore: Owners and admins can archive a company from the overflow menu. Archived companies move to the Archived tab and retain all sources, evidence, objectives, cycles, experiments, briefs, outcomes, decisions, and audit history. Archive is a reversible soft-state transition.
Tenant isolation: Company data is not visible to any other company. Server-side authorization and row-level policies enforce this on every read and write.

Sources and evidence

Bring in files, links, notes, and exports. Review evidence before it enters the canonical record.

Add material: Upload files, paste conversations or context, add URLs, or import project exports. Content is parsed on the server and stored as candidate sources.
Sensitive-data precautions: Imported material passes through scrubbers before it becomes a candidate. Redactions are recorded with reasons. AI-generated content stays a candidate; it is never automatically promoted to evidence.
Review and approve: Candidates are reviewed in the workspace. Approving an item makes it evidence with tracked provenance; declining removes it. Contradictions are preserved rather than resolved silently.
Update or remove: Sources can be updated or removed. Related evidence retains its history; downstream cycles keep their original record with the source's state at approval time.

Existing AI projects and external workspaces

Morrow works alongside the AI workspaces you already use. Import is guided, not synchronized.

Connect existing project: The Connect existing project dialog produces a structured handoff summary you can run in ChatGPT, Claude, Gemini, or Notion, then bring the resulting export back into Morrow.
Files, exports, and pasted conversations: Upload project files or exports, or paste conversations directly. Every artifact is labeled with its author (human, AI, or unknown) and its origin tool where derivable.
Context Pack: Export your company's current canonical state as a Markdown Context Pack. Use it in your other tools without leaking evidence Morrow considered sensitive.
Limitations: Morrow does not currently provide native, real-time synchronization with Google Drive, Notion, ChatGPT projects, or Claude projects. Import is one-way and reviewed.

Decision workflow

One approved cycle at a time, with frozen thresholds and preserved history.

Objective and constraint: You approve the objective for the cycle and the single constraint that most limits progress. Both are versioned; earlier versions remain in history.
Experiment with frozen thresholds: When you approve an experiment, its success and failure thresholds are frozen. They cannot be edited after freeze; a new experiment is required to change them.
Implementation brief: Morrow generates a plain-language brief derived from the approved experiment: what to do, how to measure, and when to stop.
Outcome and decision: Outcomes are recorded against the frozen thresholds. Your final decision is recorded in Decision Memory and used to inform future cycles.

Model modes

Five user-facing modes. Choose Auto unless you want to override.

Auto: Morrow picks a mode per task based on what the request needs. Use Auto by default.
Swift: Prioritizes responsiveness for lightweight requests such as short clarifications, summaries, and quick rewrites.
Standard: Balanced default for most cycle work — diagnosis, evidence review, constraint framing.
Depth: Prioritizes careful reasoning for consequential decisions, complex trade-offs, and dense evidence.
Wide Context: For requests that need a lot of context read at once — large source packs, long histories, or multi-document reasoning.

Integrations

What Morrow can connect today, and how each connection behaves.

Google sign-in: Available at sign-in. Grants only the identity needed to authenticate; no Drive or Gmail access is requested.
Stripe billing: Powers checkout and the customer portal. Stripe webhook events are the source of truth for subscription status; entitlements refresh from the webhook, not from client requests.
HubSpot and PostHog: Available on eligible plans. Connections use OAuth or API keys held server-side; disconnecting stops future synchronization and preserves prior records.
Historical records after disconnect: Disconnecting an integration never deletes what Morrow already ingested. Records remain visible to the company and continue to carry their original provenance.

Billing and plans

Plans and pricing are managed on the Pricing page. Documentation describes behavior only.

Plans and upgrade: Current plan names and prices are listed on the Pricing page. Upgrades and downgrades go through Stripe Checkout and the customer portal.
Entitlements: Entitlements refresh from Stripe webhook events. If Stripe reports a plan change, Morrow reflects it on the next authorized request.
Payment failure: A failed invoice moves the subscription to past_due and prompts you through the portal. Continued failure follows Stripe's dunning until the subscription is cancelled.
Cancellation and retention: Cancelling stops future billing at the end of the current period. Your data remains available for the retention window described in your plan; you can export or delete at any time.

Exports and portability

Take your work with you at any time.

Decision Memo: Every completed cycle can be exported as a Decision Memo suitable for sharing internally or with partners.
Context Pack: Export the company's current canonical state as Markdown to seed another tool or preserve a snapshot.
Company export: Owners can request a full export of the company's sources, evidence, objectives, cycles, experiments, briefs, outcomes, decisions, and audit history.
What is not exported: Internal model reasoning, hidden prompts, and Morrow's routing decisions are not part of any export.

Security, privacy, and data controls

Implemented controls, described at a level accurate to what the product actually does.

Authentication and isolation: Google or email sign-in with confirmation. Company data is isolated at the database level; every server request re-checks membership on the company it targets.
Credentials and webhooks: Provider keys are held server-side and never bundled into the client. Every incoming webhook is signature-verified before it can change state.
Sensitive content and imported AI material: Pasted and uploaded content passes through scrubbers before it enters the candidate record. AI-generated material is never automatically treated as confirmed evidence.
Deletion and audit: You can request deletion of a company; the request is recorded in a peppered ledger and processed on the retention schedule. Approvals, deletions, and integration changes are captured in audit history.
What we do not claim: Morrow does not hold SOC 2, ISO 27001, HIPAA, or PCI certification, and does not offer zero-knowledge or end-to-end encryption. We describe what is implemented rather than what is certified.

Reliability and transparency

How Morrow behaves when a request fails, and what it will never do quietly.

Structured-output validation: Every task that returns a structured result is validated before it is written. Invalid output is retried safely rather than persisted.
Safe retry and preserved work: Idempotency keys guard against duplicate side effects. If a model or provider call fails, your in-progress work is preserved and you are told what happened.
No fake success: Morrow will not report a request as succeeded if it did not. Failures surface truthfully, with a plain-language reason.
Historical versioning: Objectives, constraints, experiments, briefs, and decisions are versioned. Earlier versions remain readable so you can audit how a cycle evolved.

Technical overview

A public-safe summary for technical evaluators, partners, and investors.

Architecture: Chat-first React frontend on TanStack Start; server functions and server routes run in a modern managed runtime. Postgres is the system of record with row-level authorization on every company-scoped table.
Provider-neutral AI: A single orchestrator abstracts over multiple providers with structured-output validation, safe retry, and an emergency provider disable. Model choice and routing rules are not published; user-facing mode names are.
Tenant isolation and auditability: Company data is isolated at the database policy level. Approvals, deletions, and integration changes are recorded in audit history.
Portability and reliability: Companies can be exported as structured data plus Markdown. Public integrations use signed webhooks; historical records survive disconnection.
Due-diligence contact: For a detailed technical or security conversation, email support@meetmorrow.ai and mention that it is a technical or evaluation request.

Developer surface

What is and is not available for programmatic use.

Public API: Morrow does not currently expose a general public API or SDK. Internal server actions are not public integration points and may change without notice.
Webhooks (inbound): Stripe webhooks are consumed at Morrow's own endpoint using the signing secret you configure. There is no outbound webhook system today.
Exports as integration surface: The supported way to move Morrow data into another system is via Company export, Decision Memo, and Context Pack.
Requesting integrations: If you have a specific integration in mind, email support@meetmorrow.ai. We prioritize requests against real founder use.

Feedback, support, and contact

How to reach us, and what to expect.

Support email: support@meetmorrow.ai is the general support and evaluation channel. We do not publish other addresses as support endpoints.
Contact form: The Contact page routes messages to support. Your reply-to is set to the address you provide.
Feedback forms: Signed-in and signed-out feedback forms both deliver to support. Technical details are attached only when you check the technical-details box. Morrow never attaches your conversation content to a feedback message.
Response expectations: We read every message. We do not publish a guaranteed response time; urgent operational issues should mention that in the subject.

Frequently asked questions

Short answers to the questions we hear most often.

Is Morrow a general-purpose chatbot?: No. Morrow is a decision system for founders. The chat is the surface; the value is the structured cycle underneath.
Does Morrow replace my AI workspace?: No. Morrow works alongside ChatGPT, Claude, Gemini, and Notion. Bring context in with Connect existing project; keep using your other tools for what they are good at.
Does imported AI content become evidence automatically?: No. Imported material is a candidate until you review and approve it. AI-generated content is labeled as such.
Can Morrow make decisions for me?: No. Morrow proposes; you approve. Objectives, constraints, experiments, and final decisions are only recorded after your explicit approval.
Which mode should I use?: Auto by default. Use Swift for quick, lightweight requests, Standard for typical cycle work, Depth for consequential decisions, and Wide Context when a lot of context needs to be read at once.
Can I export or delete my data?: Yes. Company export, Decision Memos, and Context Pack are available at any time. Owners can request full company deletion.
Does Morrow expose one company's data to another?: No. Company data is isolated at the database policy level and re-checked on every server request.
How do I contact support?: Email support@meetmorrow.ai, use the Contact page, or use the Feedback form in the sidebar.

Continue