ra·tion·al adjective Reasoned through formal argument; explicit in its premises and inference rules such that decisions and conclusions can be explained and openly examined.
Your name is on the conclusion. SMEme checks each matter against the procedure you authorized, so you can approve the decision knowing exactly why it holds.
Build your procedure with the AI you already use. SMEme turns your policies, research, and judgment into a decision tree you approve, then compiles it into formal logic, not just vibe-coded rules, not just a flowchart. This decision tree is an explicit and easily read and maintained artifact you 100% own and control. Your AI brings in the facts from your documents and tools. SMEme shows what the facts determine, which specific sets of facts were decisive, and what facts would have to change to reach another conclusion. You use these tools for logical analysis, counter-argument and negotiation analysis, and strategic decision support. Built for the attorney, the tax adviser, and anyone else who re-checks every AI answer because their name is on the conclusion.
Works with any MCP client, like Claude ChatGPT
Works with any MCP-enabled workflow framework, like Zapier n8n LangGraph
Build a decision tree with the AI you already use, then run it on a real case.
Free to start. No credit card. The free tier includes 3 decision trees. Upgrade to Pro later if you need more. See pricing
Three prompts. Paste each one into Claude or ChatGPT and let it lead.
Connect me to SMEme over MCP. The instructions are at smeme.ai. Walk me through adding it as a connector in this app, step by step. When the sign-in tab opens I'll sign in and come back. Then confirm you can see SMEme's tools.
Build a SMEme decision tree for a decision I sign off on repeatedly: [name it, e.g. approving a new vendor]. Load SMEme's design guidance and follow it. Interview me, then stop for my approval three times: the outcomes, the questions, then the full outline. Ask me where the answer to each question comes from. When I approve, save it as a draft and tell me how to Deploy it.
My SMEme draft is saved. Walk me through Deploy and switching Listed on in the SMEme editor, step by step.
Longer prompts, including one that finds the decision for you, in the full quick start.
You ask in plain language. Your agent picks the SMEme tool. These prompts are for a decision tree you have deployed; fill in the blanks with your own tree, case, and sources. The video is a demo of one fictional matter.
smeme_reasoning_listWhich SMEme decision trees do I have deployed? Give me the name and a one-line description of each.
smeme_reasoning_evaluateRun my [decision tree name] SMEme decision tree on this case: [describe it, or attach the file]. First list the other apps and connectors you can already read. Then use the sources I name: [attached files / my mailbox / this drive / our CRM]. Gather the evidence, cite where each answer came from, and ask me when you can't find one. Then tell me the outcome and which answers decided it.
smeme_reasoning_what_ifSame case, but suppose [one fact were different]. Does the outcome change? Show me what's different from the baseline.
smeme_reasoning_how_to_reachWhat's the smallest change in the facts that would get this matter to [the other outcome]? List each fact that would have to change and what evidence would support it.
How to name your sources and how a run proceeds: Run a tree. The tool names are what your agent calls behind the scenes. Connector steps are in the MCP setup guide.
Use cases
Your AI reads these apps through its own connectors. SMEme applies the decision tree.
“Run my cross-border withholding decision tree on the Party D royalty. It is in Salesforce as matter 123. Read the agreement and the W-8BEN-E in OneDrive, the intake memo, and the Outlook mail.”
“Run my medical necessity decision tree on this order. Read the chart in athenahealth, the clinical note in Google Drive, and the payer policy in Gmail.”
“Run my suitability decision tree on this recommendation. Read the Salesforce client profile, the product disclosure in Google Drive, and the compliance note in Gmail.”
“Run my client acceptance decision tree on this prospective client. Read the Gmail thread, the intake sheet in Google Drive, and the Salesforce matter record.”
“Run my conflict check decision tree on this name. Look through the Salesforce matter list, the Teams channel, and recent Outlook mail.”
“Before today’s call, run my client acceptance decision tree. Gather the Google Calendar invite, the Gmail thread, and the notes in Google Chat.”
“A new Outlook message says the ownership changed. Run my client acceptance decision tree again with it and the chart in OneDrive.”
“Run my alert triage decision tree on this incident. Pull the alerts from Splunk and the sign-in history from Sentinel, and check the Teams channel for a change that explains it.”
“Run my release decision tree on this pull request. Read the diff, the linked issues, and what we said in Slack.”
“Run my scope decision tree on this request. Read the spec and the related project work.”
Control doesn’t come from monitoring, it comes from reasoned design.
With help from your AI assistant, you build a decision tree from your own policies, saved prompts, and best practices you’ve already approved. In decision tree format you can read every question, answer option, and branching condition. SMEme then validates the tree’s structure and you deploy it. SMEme compiles the approved tree into formal constraints, and a rational AI constraint solver automates the reasoning strictly over those constraints and whatever facts or assumptions you explicitly submit. That is control by design, not by monitoring drift or error rates.
This goes beyond getting a conclusion or simply walking a flowchart. SMEme reasons over a constraint space you authorized, and can provide logical analysis for decision support and negotiation simulation. SMEme’s analysis tools provide several practical ways to ask more about your matter. You can see which minimal set of facts decide the outcome, which open questions could possibly affect the outcome, and which change would lead to a different result, so your effort goes where it matters. Outcomes follow logically from your procedure and the facts you admit, not from a probabilistically generated guess. They are easily explained, and every fact, inference, and outcome is directly auditable to your own decision procedure. This is something no LLM can do.
If a colleague or counterparty reads a rule, answer option, or branching condition differently, that difference points directly to a specific natural-language value that can be edited and to a tree that can be expanded with new questions, branching conditions, and so forth. There is no black box to fine-tune, and no prompt to engineer, only explicit natural language rules coordinated across an easily diagrammed decision tree. SMEme cannot tell you your decision procedure is right, but it gives you precise control over that procedure and makes the automated reasoning over it transparent and fully deterministic.
The decision trees you author are yours. You can download them as a shareable file that other SMEme users can upload, and you can delete them completely from our system at will. We do not sell your decision trees or use them to train or fine-tune any model, and we use them only to run the service for you. Read more in our Sovereignty and AI Principles, and see our Privacy Policy.