Riley operates the tools you already use, watches for what slips, drafts the fix, waits for your approval, verifies it landed, and signs a receipt anyone can check. Two layers in every plan: it does the work, and it proves the work. Most AI shows you a chat log. Riley shows you a signed record of what actually got done.
Every plan includes both layers. Riley operates your tools end to end, and it proves what it did. That second layer is the part most AI skips.
Riley works where you already work. It reads your connected tools, drafts the work, and does it, with the same tools, memory, and rules everywhere. Anything that writes outside your workspace waits for your approval first.
Riley watches for what needs doing, turns problems into fix cards, and after you approve, verifies the result and records it in a signed receipt anyone with the link can check. Every line is checked against the connected account, so it is proof, not a claim.
Riley is built for agencies, businesses, ecommerce, SaaS and B2B, AI automation agencies, consultants, and creators. No audience is the default. See what Riley runs for your kind of business.
A one-star review sits nine days. A broken page loses sales all month. A lead goes quiet. You find out too late. Riley watches the accounts and tools you connect around the clock, so the small slips hit your radar early. Reviews and listings are one thing it watches; websites, leads, and stores are others.
A flag you still have to act on is work back on your plate. So Riley does not just point at the problem, it writes the fix in the client's voice: the review reply, the corrected listing detail, the answer to a question. Each lands as a card in your approvals queue showing exactly what will change and on which client. You approve, edit, or skip. Riley drafts, you decide.
Being found is one of the outcomes Riley runs, and it matters more every month. You want to be the name that comes up when someone searches your service, and now when someone asks an assistant for a recommendation. The signals that decide it are ones Riley already watches: complete listings, answered reviews, the questions real customers ask. Same loop, pointed at getting found.
Every Friday you forward one receipt per client: what Riley found, what you approved, what went live, and what is still waiting on you. Every line is checked against the connected account, so it is proof, not a claim. When Riley could not confirm something, it says so plainly instead of padding the list. The honest receipt is the one a client believes.
Riley watches, catches, and drafts on its own. Anything that touches a client account waits for you. No surprises, no apology calls.
Riley keeps a structured memory of your voice, your services, staff names, and the things never to mention, and applies it to every draft. Manage many businesses and each one gets its own separate memory. The edits you used to make every week, you stop making, and the longer Riley runs, the less it has to ask. You can read and edit the memory any time.
The same memory and the same approval rules on every surface. AI clients connect through MCP. Slack and Teams are notification and delivery channels for alerts, approvals, and receipts. Statuses are honest: verified, beta, or where supported.
The full product in your browser, no setup beyond signing in. This is the control center and the source of truth.
The raw MCP endpoint for anything that speaks the protocol, with per-workspace keys.
Verified. Connect Riley as an MCP server and work from Claude Desktop.
Verified. Add Riley with one command and use it from the terminal.
Verified. Add Riley as an MCP server in Cursor settings.
Works where ChatGPT supports custom MCP connectors on your ChatGPT plan.
Notification and delivery channel: risk alerts, approve or deny buttons, and receipts land in one channel you choose. The dashboard stays the source of truth.
Notification and delivery channel in testing: receipts and approval buttons in Teams. Setup is guided by the Riley team.
Register your own MCP server or import an OpenAPI spec; Riley uses them under the same rules.
Keys are minted per workspace and shown once. Any client that speaks MCP can connect, including Windsurf, Gemini CLI, GitHub Copilot, and more. Use Riley anywhere.
Turn on an outcome and Riley runs it on a schedule, writes what it did to the ledger, and still queues any external write for your approval inside each run.
Recurring work runs without a reminder from you, and each run records its time and result.
Everything consequential lands as a fix card or a draft. You approve, edit, or skip before anything ships.
Every run writes to the outcome ledger the moment it completes, ready for the weekly receipt.
Reviews, website QA, lead logging, brand-voice drafts, and more run today. Deeper workflows are labeled honestly as they roll out. See the outcomes.
Riley can run assumption, evidence, and risk checks on its own proposed actions and show them on the approval card. This is Skeptic Mode: built and in beta, off by default, enabled on request per deployment.
When enabled, Riley surfaces what it assumed, the evidence behind a proposed action, and the risk if it is wrong, right on the card you approve. What is genuinely live today regardless of Skeptic Mode: Riley already asks for approval before any external action, and turns approved work into signed receipts you can verify.
Buy the evidence, not the badge, before you spend money on someone's say-so.
Rewrite a strong-sounding claim in plain words and see if you would still buy it.
Argue the strongest opposing case before you commit, so you choose with your eyes open.
Treat an AI answer as a draft to check, not a fact to act on, until you see the source.
Name the wishful part out loud before it quietly drives the decision.
These five reflective checks are a thinking framework you can read and apply today. Learn them in the Academy.
Riley runs the same watch, catch, fix, verify, receipt loop across every outcome you switch on. Start from the dashboard, then use Riley where you work, and everything stays in one control center.
Approval on writes is the default, not a setting you have to find. Keys are scoped and revocable, and every action is on the record.
Anything that writes outside your workspace waits in your approvals queue, from every surface Riley works on.
Keys are minted per workspace, shown once, and revocable in Settings. Each connection is scoped to what you grant.
Every completed action lands in one ledger the moment it happens, so there is always an answer to what ran and when.
Read the full posture, data handling, and connection model on the security page. See security.
For teams that answer for their AI use, Riley will assemble AI activity across surfaces into a usage receipt: what was asked, what shipped, and who approved it. This usage receipt is coming soon and labeled that way. What is live today: requests from any connected AI surface pass through the same approval gate, and every gated action is recorded with the surface it came from.
Short, plain lessons that get you from signup to your first receipt, then deeper: how approvals work, what receipts prove, and the judgment checks you can use today. No jargon, no filler.
Open the AcademyWork that fits no preset starts with the proof loop and read-only watching, then gets shaped with the Riley team. We scope which parts are live today and which need building, with approvals and receipts from day one. Nothing is promised as live until it is.
Explore Custom SolutionsEvery plan includes the AI Work ledger, receipts, and Shield rules. Get early access, then pick the plan that fits how you work.
Plans are billed monthly, a flat price each. Included usage; no overage without approval, no surprise overages, no setup cost. External AI/provider costs are separate unless explicitly routed through Riley.
Put Riley on your hardest-to-keep clients this week and let the first Friday receipt do the talking. Get early access, run the loop on your own accounts. No card.