what you get, in writing

Service Description, in plain English.

Sub-processors, in plain English.

Sub-processors, in plain English.

Exhibit A to the Cloud Service Agreement. This Exhibit forms part of each Order Form under the Framework Terms between Wysdym Inc. (“Provider”) and Customer. It is maintained as a Linked Page and changes only as Additional Term 9 on the Cover Page provides. Terms defined in the Standard Terms or on the Cover Page have the same meaning here.

 

“Customer Agent” means an AI agent or other software client that Customer or its Users connect to the Cloud Service through the Gateway, such as Claude, ChatGPT or Cursor. “Output” means the answers, drafts, pages, Findings and proposed actions the Cloud Service generates for Customer. “Findings” means patterns the Cloud Service detects in how Customer’s agents perform, each with supporting evidence. “Plays” means workflows Customer sets up in the Cloud Service, each with a trigger (a schedule, a Finding or a manual start), the steps to take and its approval settings.

Exhibit A · Version 1.2 · November 17, 2026

five parts, one door

1. What the Cloud Service is

The wysdym platform is an operating layer for a company’s go-to-market (GTM) agents. It gives every agent Customer uses the same grounding in Customer’s own knowledge, the same limits and the same review of what it does. It has five parts, connected through a sixth, the Gateway:

 

  • Cortex: Customer’s shared knowledge and memory. The Cloud Service reads the sources Customer connects or uploads and writes structured Cortex pages about Customer’s company, accounts, products and market, citing a source for each statement about the company. A person reviews changes to Cortex page text before they take effect. Entities extracted from sources with high confidence, such as companies, people and products, are added to Cortex automatically; others are held for review.

  • Skills: a library of GTM capabilities, such as account research, lead qualification, prioritization and drafting, that Users and Customer Agents can run.

  • Governance: identity, permissions and approval queues for each agent, so that no agent can change a connected system without approval.

  • Observe: objectives Customer sets, and reports on how well its agents meet them, with quoted evidence.

  • Operator: an in-app assistant that answers from Cortex and proposes actions.

  • Gateway: the single connection point. Customer Agents connect in through a Model Context Protocol (MCP) server at mcp.wysdym.ai, and the Cloud Service connects out to Customer’s systems of record.

 

Users sign in at app.wysdym.ai through Provider’s authentication provider, which supports single sign-on and multi-factor authentication.

what it reads, and what it doesn’t

2. Sources the Cloud Service reads

2.1 Connected systems. Customer may connect the systems of record that the Documentation lists as available; Provider adds connectors over time. The Documentation also says which connectors support write access. Connections are made through Provider’s integration provider, which holds the access credentials.

 

2.2 Uploads. Users may upload files to Customer’s knowledge base.

 

2.3 Websites. Customer may add its own website and other URLs as sources, which the Cloud Service reads through a web-crawling provider. Some Skills also run web searches. Searches that contain an email address or phone number are refused. Customer will use website sources and web searches only for its own GTM research, about its own company, its competitors, its market and industry, and its accounts and prospects, and not to collect web content in bulk or for any other purpose.

 

2.4 What it reads, and what it does not. From each connected system, the Cloud Service reads the content Customer authorizes, such as records, documents, messages, call transcripts and meeting notes; the Documentation describes what each connector reads. The Cloud Service does not download call or meeting audio or video. It does not scan sources for Prohibited Data, and its classification and redaction features are not a safeguard for that data.

 

2.5 How content is used. Call and CRM content is used to write Cortex pages; pages describe it as patterns rather than quoting it, and the source text is not stored in Cortex or the search index. As Provider releases the features, the Cloud Service may also use call and CRM content, for Customer only, to identify signals about Customer’s accounts and deals, such as churn risk, growth and deal risk, and to trigger the Plays Customer sets up. Copies of the source text remain only in the Gateway logs that Section 3.4 describes, with model providers for the retention periods in Annex 3 of the DPA, and in encrypted database backups until they expire on the rotation in Annex 2 of the DPA. Cortex page text is indexed for search once a person approves it; other Cortex data is indexed when it is added.

nothing written without approval

3. Connected systems and write access

3.1 Customer sets the permissions for each system it connects, including whether the Cloud Service may write to it, and Customer’s administrators can change them at any time.

 

3.2 The Cloud Service sends a change to a connected system only where write access is on and either a Customer administrator has approved that change in the Cloud Service’s approval queue or a rule for automatic approval that Customer has turned on under Section 3.3 approves it.

 

3.3 At the date of this Exhibit, the Cloud Service offers no rule that approves changes to connected systems automatically. If Provider later releases one and Customer turns it on, Customer is responsible for the changes it approves as if a Customer administrator had approved each one.

 

3.4 Agent actions and Gateway calls are logged, and the content of Gateway calls is kept, as Annex 2 of the DPA describes.

your agents, your credentials

4. Customer Agents and third-party products

4.1 A Customer Agent that connects with credentials Customer or its Users issued acts on Customer’s behalf, and Customer is responsible for it as it is for its Users.

 

4.2 Information that a Customer Agent or User retrieves from the Cloud Service leaves Provider’s control. Its use and disclosure by the agent’s provider, and by any third-party product Customer connects, are governed by Customer’s agreements with those providers, and Provider is not responsible for them.

check before you rely on it

5. AI models and Output

5.1 Models used. The AI features run on the model providers listed in Annex 3 of the DPA, under commercial terms that forbid them from training on Customer Content. Model calls are traced for debugging as Annex 3 describes. The sub-processor list at https://www.wysdym.ai/subprocessors names every provider.

 

5.2 Output can be wrong. Output is generated by AI models from Customer Content and public sources and may be inaccurate, incomplete or out of date, even when it cites a source. Customer is responsible for reviewing Output before relying on it, and for approving any action before it is taken.

 

5.3 Research and ranking of people. Some Skills research companies and people, or rank and qualify prospects. That Output is a suggestion for Customer to review, not a decision. Customer will not use Output to make decisions with legal or similarly significant effects on individuals, for example in employment, credit or housing, without meaningful human review.

 

5.4 Ownership. Output is Customer Content under Additional Term 1 on the Cover Page, and Customer owns it to the extent the law allows. Similar Output may be generated for other customers from their own content.

how usage is counted

6. AI usage, metering and limits

6.1 AI credits. AI usage is measured in AI credits: 1 AI credit for every 1,000 model tokens processed for Customer (input and output tokens together) by Skill runs, Operator and other AI features. Ingestion, meaning reading Customer’s sources, extracting facts, describing images and writing Cortex pages from them, is included in the Cloud Service Fees and uses no AI credits, within the daily ingestion limit in Section 6.5. Web pages read and web searches run are also included, subject to Section 6.5. Provider may change the rate under Additional Term 5 on the Cover Page.

 

6.2 Allowance and overage. Each Order Form states a monthly AI credit allowance. A month is a calendar month (UTC). Model tokens are counted as Provider’s model providers report them, including cached input tokens. A call that fails is counted only to the extent the model provider reports tokens for it. Unused AI credits do not carry over to the next month. AI credits used above the allowance in a month are overage, charged at the overage rate in the Order Form and invoiced monthly in arrears.

 

6.3 Notice. Provider will tell Customer’s administrator by email when Customer’s use in a month reaches 80% and 100% of the allowance. After either email, Customer may ask Provider to stop use at the allowance for the rest of that month instead of incurring overage. Provider will then throttle AI features within 1 business day of the request and will charge no overage for use after that business day.

 

6.4 Measurement. Provider’s usage records are the measure of AI credits used. On request, Provider will give Customer a usage report for any invoiced month. Customer may dispute an overage charge under Section 4.4 of the Standard Terms.

 

6.5 Operational limits. Separately from the allowance, Provider applies limits per tenant on request rates to the Gateway and on daily ingestion model usage, and limits web reading during onboarding, to protect the Cloud Service, as described in the Documentation. Provider may throttle use above those limits and will tell Customer when it does.

not part of the service until released

7. Features not yet generally available

Features Provider marks as beta, preview or early access are Beta Products under Section 12.7 of the Standard Terms. Features not yet released, such as scheduled Plays and automatic approval of changes to connected systems, are not part of the Cloud Service until Provider releases them.

your rights, your notices

8. Customer responsibilities for data

Customer is responsible for having the rights, notices and consents the law requires for the Customer Content it submits or connects, including:

 

  • consent to record and transcribe calls and meetings that reach the Cloud Service through a connected system, including where every party to a conversation must consent;

  • the right to have the Cloud Service read each website or URL Customer adds, consistent with that site’s terms; and

  • not submitting or connecting Prohibited Data.

hosted in the United States

9. Where data is processed

Customer Content is hosted, stored and processed as Section 7.1 of the DPA describes for Customer Personal Information.

take your data with you

10. Export

Users can download uploaded files and export Observe reports as CSV from the Cloud Service. Provider exports Cortex pages as Markdown files, and other Customer Content in a commonly used machine-readable format, on written request and without charge. Export and deletion at the end of an Order Form follow Additional Term 4 on the Cover Page and Section 11 of the DPA.

The short version -

This page says what the wysdym Cloud Service does, what it reads, how writes are approved, how AI credits are counted and what isn’t released yet. Your Order Form points here. Questions go to legal@wysdym.ai.