Sally documentation

Sally is API-first project management, CRM, marketing automation, and commerce in one workspace — built so humans and AI agents work from the same brand memory, under one permission and approval model. Everything Sally does is available three ways: a human web UI, a real HTTP API, and a hosted MCP endpoint for agents. Sally is public at usesally.com and the app runs at app.usesally.com.

This page explains what Sally is, what makes it different, who it's for, and how the pieces share one context. Every section links to a deeper page.

A brand overview with the Getting Started onboarding checklist — the first thing you see in a new brand.

What Sally is

Teams increasingly work alongside AI agents, but agents are usually blind (no shared context) and dangerous (raw access, no guardrails). Sally addresses both. It's the shared operating layer where an agent — Claude, ChatGPT, or your own runtime — reads your brand memory and live context before acting, then proposes actions that Sally governs: safe, reversible work runs automatically, while risky actions (mass sends, publishing, deletions, exports, deploys) are held for human approval with full evidence and an audit trail.

Around that governance sits a genuinely useful product suite, all sharing the same contacts, context, and permissions:

Projects & tasks — Kanban boards, priorities, labels, checklists, comments, blockers, and an agent handoff/resume workflow. See Projects & tasks.
Timesheets — per-user time entries, billable flags, two-stage invoicing, and cross-project reporting. See Timesheets.
CRM — organizations, people, deals, activities, and follow-ups, plus a cross-brand relationship graph. See CRM.
Marketing automation — contacts, consent, audiences, forms, a visual email builder, campaign flows, and human-gated sends. See Marketing.
Commerce — Shopify and WooCommerce connectors that turn store orders into contacts, CRM people, and lifetime-value profiles. See Connect your store.
Memory — a typed, searchable "brand brain" of facts, decisions, constraints, and brand voice that agents read before acting. See Memory.

The five differentiators

1. One permission model for humans and agents. Every human, API key, MCP key, and agent identity resolves to a Sally account and is bounded by the same account → customer-account → brand (workspace) → project role and scope checks. An agent's reach is never broader than the credential behind it. See Core concepts.

2. Safe autonomy (Control Center). Agents submit proposals; per-brand trust policies decide auto-execute versus approval. Safe actions (memory writes, notes, tasks, drafts, handoffs) run automatically; risky actions (email.send, publish.external, record.delete, data.export, code.deploy, bulk updates) are quarantined for human approval. Blockers, approvals, and normalized evidence turn "an agent did something" into an auditable, reversible flow. See Governance.

3. Brand memory and context, agent-native. Typed operational memory (facts, decisions, constraints, preferences, brand voice) with semantic search, plus a single sally.context call that returns verified memory, active projects, open deals, overdue follow-ups, and standing instructions. Agents ground themselves before planning or acting. See Memory.

4. Agent-first surface (hosted MCP). A first-party, stateless hosted MCP server exposes the full product to any MCP-capable agent, reachable as a one-click OAuth connector in Claude and ChatGPT. The principle: you shouldn't have to use Sally to use Sally. See Connect your agent.

5. A converged suite that shares one context. Projects, timesheets, CRM, marketing, and commerce are one system, not bolted-together apps. A store order can become a marketing contact and a CRM person and a lifetime-value profile and trigger a loyalty campaign — and an agent can drive all of it under governance.

The Control Center with a pending proposal, its actions, and evidence — The Control Center — a pending agent proposal with the actions it would run and the evidence behind it, held for human review.

Who Sally is for

Agencies and studios that run client work as projects, bill by timesheet, and want a lightweight CRM and marketing engine sharing the same client records.
Brands and small teams that run marketing and (optionally) an online store, and want customer context, campaigns, and loyalty in one place.
Builders working with AI agents who want a governed, API-first backend their agents can act in safely — reading shared memory and proposing actions instead of holding raw credentials.

Sally is multi-tenant by brand (workspace): pricing gates the number of brands, not seats or features. A customer account can own several brands, each with its own projects, CRM, and marketing. See Plans & billing.

How the pieces fit together

The suite is unified by three things shared across every module:

One set of records. Projects can link to a CRM organization; deals can bridge to a project; a marketing contact can link to a CRM person; a store customer flows into all of them. There's no re-keying between modules.
One memory and context layer. Memory items link to projects, CRM records, and campaigns. Agents call sally.context once to orient, and project.runtime_context or task.resume_context to pick up specific work.
One governance model. Whether a human clicks a button or an agent calls a tool, the same role checks and the same Control Center proposal/approval flow apply.

For agents, each capability is exposed as an MCP tool in dotted form (for example task.handoff.create, crm.person.create, marketing.message.propose_send); the connector surfaces the same tools underscore-prefixed (mcp__sally__task_handoff_create). Start every agent session with sally.context.

Start here

Getting started — create your account and first brand, invite your team, connect your agent.
Core concepts — accounts, customer accounts, brands, projects, the role model, identity tiers, and governance.
Connect your agent — give any MCP-capable agent access to Sally.

Working in Sally

Platform and integration

Governance — Control Center: proposals, trust policies, approvals, blockers, evidence.
HTTP API — the same surface over HTTP, with auth paths and resource routes.
Security — passwords, sessions, 2FA, SAML SSO, key policies, and audit logs.
Plans & billing — plans, brand limits, trials, and signup.

Roadmap

Some capabilities are in progress and not yet available. We'll describe them as shipped only when they are:

Additional send channels. Marketing sending is currently email only (SMTP / AWS SES). SMS, WhatsApp, and Telegram are planned.
More executable opportunities. Of the next-best-action wedges, only meeting-to-memory executes today via opportunity.execute; the others are recommend-only. See Opportunities.
External work-item sync. Sally tasks are the wired system today; Jira, Linear, and GitHub sync are planned.
Passkey sign-in. Passkey (WebAuthn) registration is available; passkey login is still being finished. See Security.