These terms are an agreement between you and Workgraph, Inc (“we”, “us”) and govern your use of the service at useroom.ai. Last updated 12 October 2026. They include a binding arbitration clause and a class action waiver. See “Disputes” below.
By creating an account, continuing past sign-in or using the service, you agree to these terms and to the privacy notice. If you use room for a company or other organization, you confirm that you have authority to bind it. “You” then means that organization and its users. room is for business use only. It is not offered to consumers, and you must be at least 18.
room helps companies prepare claim-based briefs from selected evidence, find relevant fits and make introductions after both sides accept. It also supports existing governed Room workflows. The service is designed so that disclosures between teams require authorization from the team that owns the data and are recorded. We do not warrant the accuracy of answers produced by model providers or by the service. We may change, suspend or discontinue features at any time. We give notice where practical.
Where email onboarding is enabled, you may join by intentionally emailing a@useroom.ai or aura@useroom.ai without an invitation. Room may create your person account and public contact page from your sender details, and you may choose topics or accept, decline and ask about requests by authenticated email reply. Forwarding a message or Cc’ing Room instructs Room to contact the parsed original sender or addressed recipients automatically; email-only third-party replies use neutral branding, with daily limits on Room-initiated contact and bounds on ongoing conversations. Room may queue contact for a later day when a daily limit is reached. Creating the account by email does not record agreement to these terms: your first interactive sign-in, through Continue to Room with the linked notice, records acceptance. This email-onboarding rule takes precedence over the general account-creation acceptance wording above.
You must use a work email you control and give accurate company information. You are responsible for all activity under your account, for the people you invite to your team and for credentials you create for agents. Keep credentials secret; revoke them if they may have leaked, and tell us promptly about any unauthorized use. We may suspend accounts that misrepresent a company or misuse the service.
You keep ownership of what you contribute. You grant us a worldwide, non-exclusive, royalty-free license to access, obtain, use, copy, store, transmit, process, display and modify your content as needed to operate, secure and support the service. The license covers content in whatever form you provide it and any access you give us to it. Our personnel and service providers may exercise it on our behalf. Model processing, staff review and matching follow the permissions recorded in the product; saving content alone does not grant them.
Submitting a brief is your approval for teams and people you’re matched with to see it as written. You separately decide whether to accept each introduction. An approval you record is your authorization for that exact use, and you are responsible for the content you approve.
You represent that you have all rights, consents and notices needed to contribute your content, including personal data about other people in connected tools, and to authorize the uses above. Do not contribute special categories of personal data, health information, payment card or bank account numbers, government identification numbers, or data about children. We have no liability for content submitted in breach of this section.
When a connection is available, these scopes describe the supported imports. Your recorded permission fixes the destination and selection. Room reads from connected tools; it does not create, change or delete their content. Saving an import does not itself grant model processing, staff review or matching permission.
New automatic connections to GitHub, Notion, Google Calendar, Google Contacts, HubSpot, Stripe, Salesforce and Granola, and new Google Drive selections made from the Connections page, authorize hourly checks within the scope and limits described here. Existing one-time imports do not gain a schedule. Gmail, Microsoft 365 and Slack imports use your selected scope and refresh only when requested. Each check is bounded by the same record and size limits as the initial import. This hourly schedule is pinned to the authorizing controller and can be stopped by disconnecting. Identical imports keep their saved version; rolling date windows can change the saved content as time passes. The schedule does not grant processing, model access, matching or funding permission.
Google Docs, Sheets (first sheet), text, Markdown, CSV and JSON files, up to 10 MB each. Google’s selected-file permission permits edits; Room only reads.
Up to 1,000 selected repositories, with up to 10 small code and documentation files per repository. This is a sample, not a complete repository export.
Text from your shared pages and child pages, imported in batches over time. Files, database exports and synced blocks are excluded. Notion may take a moment to return recently shared pages.
Accessible accounts, contacts, opportunities and opportunity contact roles, plus tasks and events created in the last 90 days including today. Standard fields include names, contact details, descriptions, relationships, dates and recorded deal status. Up to 5,000 records and 10 MB per object type per check. Custom fields and objects, leads, email bodies and attachments are excluded. Missing fields or inaccessible object types limit the context available.
Room only reads. Salesforce’s API permission can allow changes depending on your user permissions; an administrator can restrict the authorizing user to read access. Recorded deals and activities do not establish current demand, verified revenue or willingness to introduce.
Contacts: email, first and last name, company, job title and lifecycle stage. Companies: name, domain, industry, city, country and employee count. Deals: name, amount, stage, pipeline, close date and creation date. Record IDs are included. Up to 5,000 non-archived records and 10 MB per object type per check.
Charges from the last 31 days: payment and customer IDs, dates, amounts, currency, payment/refund status, description and test/live mode. Up to 250 records and 240 KB per check. Card numbers, billing details and bank details are excluded.
Your primary calendar over the previous 180 completed UTC days: titles, descriptions, dates, organizers, invitees and RSVP status. Up to 5,000 events and 10 MB per check. Read-only access; conference links and attachments excluded. An invitation or RSVP does not establish attendance or willingness to introduce.
Saved contact names, email addresses and organization fields, up to 5,000 contacts and 10 MB per check. Read-only access; phone numbers, private notes, directory and other contacts excluded. A saved entry does not establish interaction, current employment or willingness to introduce.
Messages matching a selected search query within a chosen date range of up to 31 days: from, to, cc, subject, dates and message/thread identifiers only. Read-only access. Message bodies, snippets and attachments are excluded. Interaction does not establish willingness to introduce.
Messages from a selected sender within a chosen date range of up to 31 days: participants, subject and message text. Read-only access. Attachments are excluded. Interaction does not establish willingness to introduce.
Top-level messages in a selected channel within a chosen date range of up to 31 days. Read-only access. Thread replies, files and channel membership are excluded. Participation does not establish influence or willingness to introduce.
Meeting notes and provider-generated summaries from the last 30 days that the connected Granola account can read. Available notes depend on your Granola plan and workspace access settings; some plans exclude public workspace notes. Room records the account, active workspace and note-access scopes at connection and stops importing if any change; reconnect to continue. Up to 100 meetings and 10 MB per import. Read-only access via Granola MCP. Transcripts and query tools are excluded. Only supported complete responses are saved; unsupported or incomplete responses stop the import. A meeting note does not establish attendance or willingness to introduce.
Disconnecting removes the saved acquisition credential and stops future imports. Already imported versions remain stored. Any later processing or matching requires its own current recorded permission.
When Explore is enabled, sending a question, sending a revised question or choosing Agree on question records your authorization for that exact question, your selected context and the current Explore processing agreement. No separate processing checkbox is required. Opening a page, accepting an introduction or accepting these terms alone does not authorize an exploration. Both connected company profiles must agree to the same question before their question enters the staff queue. Agreement does not start analysis; a platform operator may start one analysis under the still-current agreement when processing is available. A revision requires new agreement.
Room sends the question, both introduction-pinned shared briefs, company profile names and permitted summaries of selected context to external AI providers (a routing service and the company that runs the model), through one fixed route without fallback, with data collection denied and zero data retention requested. An external provider reads whatever is sent to it; hosted Room operators can technically access processed material. Each agreement pins its purpose, exact inputs, processing artifact and terms. Changed terms require a fresh action; earlier grants are not expanded.
The first version supports up to four structured JSON summaries per side, up to 32 KB each. It admits source-reported contract status, total-fee band, currency, term and supported technical capabilities. Customer names, exact amounts, other original fields and unsupported documents are excluded by this adapter. The shared briefs are included as written. This does not authorize analysis of other connected documents or ongoing chat messages. Agreements expire after seven days. Room funds one analysis attempt, with at most ten launches per profile per rolling day; failed analysis requires a fresh question.
Sending or agreeing shares the question with both profiles and authorizes the same proposed plan for their current team operators, including later members. Once both agree, Room staff can see the question, participants, initiating actor and processing/delivery status. This grants no staff document or plan review. Notifications link to the question or plan without including their contents.
Closing a question stops pending work and blocks plan delivery that has not occurred. Provider processing already started cannot be recalled. Each team retains a plan already received. Membership, source authority, current connection and expiry are rechecked before processing and first delivery. A plan is a proposal, not a commitment by either side or independent verification of the supplied information.
When Explore is enabled, Send question, Send revised question and Agree on question authorize the exact question, your selected immutable context versions and this versioned agreement. Both connected profiles must agree before their question enters the staff queue. Agreement does not start analysis: a platform operator may start one analysis under the still-current agreement when processing is enabled. A revision needs renewed agreement. A page visit, introduction or these terms alone grants none of this authority.
Room sends the question, both introduction-pinned shared briefs and profile names to external AI providers (a routing service and the company that runs the model), through one fixed route without fallback, with data collection denied and zero data retention requested. When document input is enabled in the exact offer, the providers also receive the complete text of every selected document; otherwise only supported structured summaries are admitted. An external provider reads whatever is sent to it; hosted Room operators can technically access processed material. Each action pins its purpose, input class, exact source versions, processing artifact and terms. Changed input availability or terms require a fresh action. Earlier agreements and credentials do not expand.
Document input supports one complete UTF-8 TXT, Markdown, CSV, TSV or valid JSON file per selected item, with up to four items per profile, 32 KB per item, 40 KB per side and 80 KB per pair including serialized input framing. Every selected item must validate in full; no silent omission or truncation. PDF, Office, binary files, invalid formats and oversized inputs are refused. Upload a correctly typed text export or a smaller explicit excerpt. A selected excerpt is analyzed as that entire excerpt, not as its larger original. No ongoing chat, unselected context, tools or external URL retrieval is authorized.
Sending or agreeing shares the question and authorizes the same proposed plan, including derived facts from selected text, to both named profiles. A personal profile's plan is accessible only to its owner. A company profile's plan is accessible to its current team operators, including later members. Original documents, source links, filenames, paths, credentials and private traces are excluded from counterpart projections. Derived facts can reveal information from your selected text; prompt instructions and output checks do not establish semantic confidentiality. Only select text you authorize for this bilateral derived plan.
Once both agree, platform staff can see the question, participants, initiating actor and processing/delivery status, with no source or plan review permission. Notifications contain links, not question or plan content. Agreements expire after seven days. Room funds one analysis attempt, at most ten launches per profile per rolling day; failed or uncertain analysis needs a fresh question and is not automatically retried. Closing stops pending work and blocks unreceived plan delivery; processing already started cannot be recalled. Each exact recipient keeps its received history under current personal-owner or company-operator access. Membership, source dependencies, current connection and expiry are rechecked before processing and first delivery. Plans are proposals, not commitments or independent verification.
An agent may exercise this profile-aware contract only under separately renewed manage_explore_documents permission for the exact profile. That permission covers these questions and bilateral derived plans, including complete selected text when enabled; existing manage_explore and matching grants keep their older scope. Read permission can still retrieve eligible historical plans. Agent access is revocable, and changing profile does not transfer source or recipient authority.
To read, assess and draft, Room sends the text it needs to external AI providers (a routing service and the company that runs the model). They receive bounded input for the purposes of the action you took or the permission you recorded. For Room’s current features (briefs, Connect, matching, Explore and understanding your context), each request uses one fixed route, with no fallback to other providers, data collection denied and zero data retention requested. Room workflows (data mapping, recurring partner Rooms and Clean Room Runs) use the providers and settings recorded in their own setup, and those may not request zero retention or a fixed route. We choose these providers and models and may change them; a change never asks you to consent again or ends a permission you recorded. An external provider reads whatever is sent to it and processes it under its own terms. The service operator and separately admitted staff reviewers can read data processed in hosted execution.
These terms cover Room’s use of external AI providers for the processing you authorize. Forwarding or Cc’ing an email to Room covers sending that email to them to read it.
Selecting Assess this text when uploading a document or saving notes, or choosing Assess text for a saved item, admits that exact item for assessment. Room extracts text in an isolated local sandbox (PDF assessment is unavailable where the host cannot support it), then sends up to 8,000 characters per item and 24,000 per request to external AI providers (a routing service and the company that runs the model), through one fixed route without fallback, with data collection denied and zero data retention requested. Each external provider reads the text sent to it; the hosted service operator can technically read processed data. PDF extraction covers at most the first 20 pages and excludes images. Scanned, encrypted or unreadable PDFs and Office files stay saved and unassessed. Coverage is shown beside the item.
The receiving profile gets derived assessment notes and may use those notes for its private next step through external AI providers. It does not receive the original document. Notes you write remain visible as stated in the request. Each assessment records the exact admission, source hash, extraction coverage, provider request and result hash. Requests use finite attempts and reread limits. Older items are not admitted automatically: Assess text renews that exact item. Extracted text remains sealed with the request, including after decline, archive or abandonment, until the item is removed. Removing an item from an editable request stops future use; it cannot recall text already sent.
Each recorded processing agreement fixes its purpose, reviewers, input scope, expiry and funding bounds. The provider and model used are recorded as facts, not terms of the agreement. When personal understanding is enabled, your successful supported upload or source selection records the current agreement for that context as part of the ordinary intake action. No separate approval panel is required. These terms alone, an ordinary page visit and previously saved context do not create or expand a processing permission. Saving without an admitted processing contract grants none. Existing Room workflows retain their own declared providers and permissions.
Personal understanding uses your admitted uploaded text and selected Google Drive context to build an internal description of your work, skills, projects, relationships, availability and intent. Each source stays with its selected personal profile. This agreement does not authorize matching, staff review, outward disclosure or introductions. Your submitted brief retains its separate sharing approval.
Room sends the bounded evidence and derived context it processes to external AI providers (a routing service and the company that runs the model), through one fixed route without fallback, with data collection denied and zero data retention requested. Each external provider reads whatever is sent to it. Processing runs on hosted, unattested infrastructure, and the service operator can technically read data processed there.
Model permission lasts up to 90 days and uses the account’s existing finite funding pool. More profiles, additional sources or renewed permissions do not reset that pool. Processing runs in bounded passes of up to 24 text units and 96,000 bytes, with up to 6,000 output tokens per pass; files up to 10 MiB can be prepared. Large selections may need multiple passes, and unfunded work waits. Unsupported files may be saved without being processed.
New versions within an admitted Drive selection may update the internal understanding. A new file selection or replacement upload records its own scope; unchanged inputs do not cause repeat processing. Previously save-only content is not automatically included. Adding context does not resume paused, expired or superseded model permission. You can pause new processing and explicitly resume under the current agreement.
The admitted scope also permits continued storage and deterministic indexing after disconnection. This storage enrollment outlasts the model permission and grants no further provider, matching or disclosure access. Disconnecting stops new acquisition; retained imports may still be used under current processing authority. Pausing model processing does not delete saved originals or completed internal profiles. Contact us to request deletion. The Google-specific limits in the privacy notice take precedence.
room suggests possible connections. It does not vet, endorse or guarantee any business, person, claim or brief, and it is not a broker, agent or party to any dealings between users. Each introduction depends on both sides accepting. You are solely responsible for your own diligence, for your communications and agreements with other users, and for any confidentiality terms you want with them.
Once you accept an introduction, the other business receives the information you approved, and after mutual acceptance it receives your identity and your messages in the shared chat. We do not control and are not responsible for what another business does with information you chose to share. To the extent permitted by law, you release us from claims arising out of disputes with other users or their conduct.
Drafts, claims, fit assessments, summaries and other output are generated partly by machine learning models and may be incomplete, out of date or wrong. Review output before you approve or rely on it. Output is not legal, financial, investment, tax or other professional advice. We do not guarantee that the service will find any match, produce any introduction, or lead to any business result. Messages with room or its agents do not create contracts or obligations beyond these terms.
room works with services we do not control, including the tools you connect, model providers, hosting, email and support chat. Your use of a connected tool is governed by your agreement with its provider. We are not responsible for third-party services, including their availability, security or handling of data you authorize us to send them.
Founding, referral and other promotional credits are awarded at our discretion under the offer’s stated rules. They have no cash value. They cannot be sold, transferred or exchanged for money. They may be used only as and when we make redemption available, and they may expire or be revoked if the offer rules are not met, the award was made in error, or the account is involved in fraud or abuse. Research allowances limit processing and do not carry over unless we say so.
If you buy a paid plan, the price and billing terms shown when you buy apply. Fees are non-refundable except where the law requires otherwise or we state otherwise in writing. You are responsible for applicable taxes. We may change prices for future periods with notice.
If you send us feedback, including ratings of drafts, matches or introductions, we may use it without restriction or obligation to you. We may collect data about how the service is used, and create aggregated or de-identified data that does not identify you or your company, to operate, secure and improve the service. These rights are subject to the Google-specific limitations in our privacy notice.
Do not upload data you are not permitted to share. Do not attempt to access another team’s data outside the product’s authorizations. Do not probe, overload or disrupt the service or get around its limits. Do not copy, reverse engineer or scrape it, or access it by automated means other than the credentials and interfaces we provide. Do not misrepresent your identity or company, use introductions to send unsolicited bulk messages, or use the service to harass, defraud or break the law. We may remove content or suspend access to protect other users, the service or us.
You may stop using the service at any time and ask us to close your account. We may suspend or terminate your access, with notice where practical, if you breach these terms, if your use creates risk or legal exposure for us or other users, if the law requires it, or if we stop offering the service. When access ends, your right to use the service ends and unused credits lapse. Content and records are handled as described in the privacy notice.
THE SERVICE, ITS OUTPUT AND ALL CREDITS ARE PROVIDED “AS IS” AND “AS AVAILABLE.” TO THE FULLEST EXTENT PERMITTED BY LAW, WE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, NON-INFRINGEMENT AND ACCURACY. WE DO NOT WARRANT THAT THE SERVICE WILL BE UNINTERRUPTED, SECURE OR ERROR-FREE, THAT CONTENT WILL NOT BE LOST, OR THAT ANY OTHER USER OR THIRD PARTY WILL ACT AS EXPECTED.
TO THE FULLEST EXTENT PERMITTED BY LAW, WE WILL NOT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, EXEMPLARY OR PUNITIVE DAMAGES, OR FOR LOST PROFITS, REVENUE, BUSINESS OPPORTUNITIES, GOODWILL OR DATA, EVEN IF WE WERE TOLD THEY WERE POSSIBLE. OUR TOTAL LIABILITY FOR ALL CLAIMS RELATING TO THE SERVICE OR THESE TERMS IS LIMITED TO THE GREATER OF THE AMOUNT YOU PAID US IN THE TWELVE MONTHS BEFORE THE EVENT GIVING RISE TO THE CLAIM AND US$100. THESE LIMITS APPLY TO ANY THEORY OF LIABILITY AND EVEN IF A REMEDY FAILS OF ITS ESSENTIAL PURPOSE. They do not limit liability that cannot be limited by law.
You will defend, indemnify and hold harmless Workgraph, Inc and its officers, employees and agents from and against third-party claims, and related losses, damages and costs including reasonable legal fees, arising from your content, your sharing and approval decisions, your dealings with other users, your breach of these terms, or your violation of law or of anyone’s rights. We may control the defense of any such claim. You will not settle a claim that affects us without our written consent.
These terms are governed by the laws of the State of Delaware and applicable United States federal law, without regard to conflict-of-laws rules. Before starting a formal dispute, contact us and give us 30 days to try to resolve it.
Any dispute that is not resolved will be settled by binding arbitration administered by the American Arbitration Association under its Commercial Arbitration Rules, before a single arbitrator, and governed by the Federal Arbitration Act. Either party may bring an individual claim in small claims court, and either party may seek injunctive relief in court to protect its intellectual property, data or systems. Any court action allowed by these terms will be brought in the state or federal courts located in Delaware, and both parties consent to their jurisdiction.
YOU AND WE MAY BRING CLAIMS ONLY IN AN INDIVIDUAL CAPACITY, NOT AS A PLAINTIFF OR CLASS MEMBER IN ANY CLASS, COLLECTIVE OR REPRESENTATIVE PROCEEDING, AND EACH PARTY WAIVES ANY RIGHT TO A JURY TRIAL. Any claim must be brought within one year after it arises, or it is permanently barred, unless the law requires a longer period.
We may update these terms. We post changes here with a new date, and for material changes we also notify account holders by email or in the product. Changes take effect when posted unless we state a later date. If you continue to use the service after a change takes effect, you accept the updated terms. If you do not agree, stop using the service and ask us to close your account.
These terms, the privacy notice and the approvals and processing agreements recorded in the product are the entire agreement between you and us about the service. If a provision is found unenforceable, it is limited to the minimum extent necessary and the rest remains in effect. Our failure to enforce a provision is not a waiver. You may not assign these terms without our consent; we may assign them in connection with a merger, acquisition or sale of assets. Neither party is liable for delays caused by events beyond its reasonable control. Nothing in these terms creates a partnership, agency or employment relationship. You confirm that you are not on a U.S. government restricted-party list or located in an embargoed country, and you will comply with applicable export and sanctions laws.
We send notices to your account email. You send notices and questions through the contact form. The sections on content, introductions, generated output, credits, feedback, disclaimers, limitation of liability, indemnity, disputes and this section survive termination.