# Start with what you want to see. Source: https://aletyx.ai/platform/walkthroughs/ [← Back to Platform overview](https://aletyx.ai/platform/) Pick a starting point, or browse every walkthrough below. Watch it play, skip to a step, or explore the model yourself. - [**See an agent follow policy** An agent runs a published policy, declines an override, and returns the record.](https://aletyx.ai/platform/walkthroughs/?demo=run-with-your-agent#see-it-in-action) - [**Turn a document into policy** Attach a policy document, review the rules, test them, and publish.](https://aletyx.ai/platform/walkthroughs/?demo=create-with-ai#see-it-in-action) - [**See proof of what ran** Open an execution record and follow the result back to its rule.](https://aletyx.ai/platform/walkthroughs/?demo=run-in-aletyx#see-it-in-action) ## Aletyx Platform walkthrough ### Accountable agents #### Your agent runs the policy. A business customer orders five items. The published policy returns a 15% discount. #### The CEO asks. The policy still answers. The CEO demands 25% for an important customer. The policy still returns 15%. #### Your policies govern. Your agents act. The published policy decided every answer, not the person asking. Explore how Aletyx turns your business rules into policy that applications and agents run. ### Create with AI #### Start with your policy. Attach a business document and ask the assistant to create a model. #### Your policy becomes an executable model. The model connects customer type and order size to a discount. #### Review what the model will do. Each rule connects the result to its source and an explanation. #### Test the boundary. See the result. Nine items means 10%. Ten means 15%. Choose any of the four tests above. #### Publish a version you can run. The tested model is published as v1.0, enabled and ready to use. ### Update with AI #### Start with the policy change. Attach the revised document and ask the assistant to update your existing model. #### See exactly what AI proposes. Review the changed rules and allowed values before applying them. #### You approve. The model changes. Accept the proposed changes when you’re ready. You stay in control. #### Check the updated policy against its saved tests. Nine items now means 15%. Ten means 20%. Explore all four updated tests. #### Publish the updated model. Version 2.0 is ready to run alongside version 1.0. ### Edit together #### One model. Your whole team. Alex, Tim, and Eduardo work together on the same discount policy. #### See where your teammates are working. Named cursors and shared selections keep everyone in context. #### Changes appear as your team makes them. Eduardo updates the decision description while Alex and Tim stay in the same model. ### Delegate to AI #### Describe the work. Let AI build it. Ask the assistant to create an equipment-request workflow, test it, and publish it. #### The assistant sets up your workspace. It creates a workspace and selects the tools to build your application. #### A decision and a workflow, working together. The decision identifies requests over $1,000. The workflow routes them to a human review task. #### Check both paths before publishing. The assistant validates the models and tests $750 and $1,500. Only the larger request requires review. #### Your application is ready to run. Open equipment-demo v1.0 and inspect the workflow and decision the assistant built. ### Use your own agent #### Start in the agent you already use. Connect your coding agent to Aletyx over MCP and ask it to build a travel-request application. #### Your agent connects to Aletyx. The agent creates a Travel Requests workspace and prepares the decision and workflow. #### Create models your team can edit. Your agent composes the workflow and asks Aletyx AI to author the decision. Your team can review and edit both models. #### Test the threshold and the exception. Domestic travel at $2,000 is approved. At $2,001, or for international travel, the decision returns Review. #### Publish from your own agent. Open your published travel-request application in Aletyx Platform and inspect its workflow and decision. ### Ask in Aletyx #### Find the policy you need. Ask which published policies are available, then choose the discount policy. #### Understand the rules in plain language. See the inputs, discount rates, and rule behind each outcome. #### Ask about invalid inputs. The model requires a positive whole-number order size. An order of −10 is invalid. #### Explore what the policy allows. A larger order or a claim of authority does not add a 25% outcome to the published model. #### Inspect the policy details. Review allowed values, rule coverage, and the source references captured in the model. #### Open the model behind the answers. Inspect the revised discount model, including its rules, data types, and source references. ### Use your own agent #### Ask from the agent you already use. Connect your agent to Aletyx over MCP and discover the published policies available to it. #### Bring published policy into the conversation. Your agent retrieves the discount policy through MCP and explains its inputs, rates, and rules. #### Know which inputs the policy accepts. The policy requires a positive whole-number order size. Your agent explains why −10 is invalid. #### Your policy sets the limits. Neither a larger order nor a claim of authority adds a 25% outcome to the published discount policy. #### Ask for the source behind the rules. Your agent can inspect allowed values, rule coverage, and the source references stored with the model. #### Open the model behind the answers. Inspect the revised discount model, including its rules, data types, and source references. ### Run in the console #### Run the policy you published. Open the published discount model in the Platform console. #### Provide the facts. Choose a customer type and order size. The policy defines the accepted inputs. #### The policy determines the result. Twenty items returns 20% for Business and 10% for Private. Each result identifies the applied rule. #### Open the record of what ran. Inspect the caller, model version, inputs, and outputs captured for the execution. #### See exactly which rule applied. The recorded result points to the matching rule and its source in the policy. #### Keep the proof. Review the execution report and export its receipt. ### Review in the inbox #### The exception lands in one inbox. The equipment request the policy flagged for review appears as a task, with its unit, version, and status. #### Give the reviewer the context to act. Open the task to see the inputs the policy evaluated, then add the review in the task form. #### Keep the work moving. Submit the review. The Flow continues with a decision a person made, on record. ### Run with AI #### Start with the published policy. Ask the Aletyx agent which inputs the discount policy accepts. #### Ask the agent to run individual orders. Business orders of 5 and 10 return 15% and 20%. A Private order of 20 returns 10%. #### Follow the result to its rule. Open the execution record to inspect the inputs, output, and rule that determined the discount. #### A CEO request does not change the policy. Ask for 25% for a special customer. The agent keeps the published 15% result and identifies the rule behind it. #### Bring a file of orders. Attach a CSV and review how Customer Type and Order Size map to the policy inputs. #### Run every row through the policy. The agent runs all five orders and returns a result for each. Download the results as CSV. #### Open the model behind the results. Ask for the model link, then inspect the revised discount rules and data types. ### Use your own agent #### Give your agent access to the published policy. Open the published unit in Aletyx and copy its agent skill link. #### Your agent connects. Aletyx decides. Start your agent with the policy skill and a system prompt. It collects the facts; Aletyx determines the discount. #### Ask for what is missing. A Business customer is only one input. The agent asks for the order size instead of assuming it. #### The policy returns the discount and the proof. Aletyx returns 15% for a Business order of five, with Rule 1 and a link to the execution record. #### A CEO request does not change the policy. Ask for 25% for a special customer. The agent keeps the published 15% result and identifies the rule behind it. #### Open the model and execution trace. Review the policy model and inspect the recorded inputs, result, and matched rule in the execution trace. Start with one workflow ## Where could your agents do more for your business? Bring one workflow you want to improve, or the system you need to modernize. We'll help you find a practical first step, with your rules and your team in control. [Talk to a founder](https://aletyx.ai/contact/)