Tejma LLC logo
DispatcherEmail SummariesLead IntakeHow We WorkTrustBook a Consultation
Tejma LLC logo
DispatcherEmail SummariesLead IntakeHow We WorkTrustBook a Consultation

Tejma Dispatcher

Privacy Policy

This Privacy Policy explains what information Tejma LLC ("Tejma", "we", "us") collects when you use Tejma Dispatcher — our message-ingestion and task-automation service — how we use it, who we share it with, how long we keep it, and the choices you have. Tejma Dispatcher exists to process your communications, so please read the "Information We Collect" section carefully: connecting a channel means the messages flowing through it, including messages written by other people, are stored and processed in your account. The service's AI processing runs on Anthropic's Claude models — under Tejma's own Anthropic account when your organization uses managed AI billing, or under your organization's account when it supplies its own API key — and the "AI Processing" section explains what reaches Anthropic in each case, and what reaches the connectors, including Tejma's own connector services, that your agents use.

Effective October 30, 2026 · Last updated October 30, 2026 · Reno, Nevada

On this page

  1. 1. Who We Are and What This Covers
  2. 2. Information We Collect
  3. 3. How We Use Information
  4. 4. AI Processing
  5. 5. Service Providers and Recipients
  6. 6. Data Retention and Deletion
  7. 7. Security
  8. 8. Your Rights and Choices
  9. 9. Email Provider Account Data
  10. 10. Children, International Transfers, and Changes

Related documents

  • Tejma Dispatcher Terms of Service
  • Tejma Dispatcher Data Processing Agreement

1. Who We Are and What This Covers

Tejma LLC is a Nevada limited liability company based in Reno, Nevada. This policy covers the Tejma Dispatcher service — its API, MCP endpoint, dashboard, and background workers. It does not cover third-party services you connect to it (your email provider, Slack, Discord, Stripe, Anthropic where you hold your own account, or connector servers you register by URL), which have their own privacy policies. Tejma's own connector services offered in the service's catalog (for example Tejma Books) are separate Tejma services with their own privacy policies; this policy describes what the dispatcher shares with them.

Our role. For business customers, the organization that connects channels and runs agents is the controller of the content processed through its account; Tejma processes that content on the organization's behalf as described here and, where applicable, under our Data Processing Agreement.

Roles in your organization. Where this policy refers to your organization's administrators, it means the members holding the roles the Terms of Service define: the billing role (plans, credit, keys), the usage role (starting and shaping AI work), and the users role (members, invitations, connector grants).

2. Information We Collect

Account information. Sign-in is handled by WorkOS AuthKit. We receive your WorkOS user identifier, email address, first and last name, profile picture URL where set, whether your email is verified, your organization membership, when you joined, and when you last signed in. We store your user identifier and organization in our database, together with your extraction and routing skills and the models they pin, your billing start date, and any Anthropic API key override encrypted; we read your name, email, and sign-in details from WorkOS when they are shown, for example in a member list. We also store the roles and the Tejma connector grants you hold, each with the identifier of the member who granted it and when, and — for invitations — the invitee's email address with the plan, roles, and connector grants the invitation carries until the invitee first signs in. We do not store your password — authentication credentials live with WorkOS.

Terms acceptance records. When you accept the Terms of Service and this policy, we record the version accepted and the time, under your user identifier. These records are append-only and are kept as evidence of consent after you leave an organization or your account is deleted.

Messages from connected channels and clients. When you or your organization connects a channel, we receive and store the communications flowing through it. For an email we store its headers (sender, recipients, copy recipients, reply-to address, subject, and date), the list of its attachments, and its body converted to plain text; since October 26, 2026 the quoted history of earlier messages is stripped from the body of new emails (forwarded emails are kept whole, and emails stored before that date keep their quoted history). For an email that continues a conversation, we also fetch from your mailbox and store a bounded plain-text transcript of the thread — the most recent messages, each truncated — which can include your own replies and messages that predate the mailbox's connection, so tasks are understood in context; we additionally record the thread's participants and subject so agents can reply into the thread. For Slack and Discord messages, delivered through Slack's Events API and Discord's gateway, we store the workspace or server, the channel, the author's display name and identifier, the time, the text, and a bounded transcript of the earlier messages in the thread. When someone replies to a message the service sent, the stored transcript is prefixed with the original message of the task that send belonged to, which may bring content from one channel into another's record. Messages submitted through MCP clients, the dashboard, or our API are stored the same way; an MCP submission also records the identifier of the sign-in token that made it.

Files and attachments. Email attachments, files shared with chat messages, and files uploaded through MCP clients or the API (up to 50 MB each) are downloaded and stored as documents on their message, in an object store keyed to your user account. Stored files are encrypted at rest by the storage provider, not additionally by the application.

Content you create. Tasks (including tasks your agents create when they pause for prerequisites), agent definitions and skill instructions, your extraction and routing skills and the models they pin, approval workflows and the evaluations they produce (including judge reasons and rejection suggestions), connector registrations and Tejma connector installs, outbound channel links, and the messages your agents send. Every successful outbound send is recorded with the channel, recipients, subject, full body text, delivery mode, the provider's message identifier, and the inbound message it answered. The service does not store a transcript of an agent's run: the model's turns and tool calls exist only in memory while the task runs and at Anthropic. What persists from a run is the task's outcome notes written by the model (including, for a completion report, a summary of where it was delivered), the descriptions of any prerequisite tasks it created, the text of messages it sent, and, for a judge, its reason.

Credentials, encrypted. Channel OAuth grants and the scopes they were granted with, connector authorization tokens, and — if your organization chooses own-key billing — your organization's and members' Anthropic API keys. A connector authorization token is your own access token for that connector, obtained when you sign in to it (through Tejma's WorkOS application for a Tejma connector, through the server's own sign-in otherwise) or pasted by you, and refreshed by the service before it expires. These are encrypted at rest with AES-256-GCM, decrypted only at the moment of use, and never returned by any API — responses carry only a has-token flag. Deleting a connector, or a seat, deletes our copy of the token; it does not revoke the authorization at the account that issued it, which you can do there. Tejma's own managed Anthropic API keys are stored the same way and are never disclosed to customers.

AI usage records. One record per AI call made on your behalf: organization, user, agent, task or message, the job (extraction, routing, approval judgement, or execution), model, attempt, the four token counts Anthropic's API reported (input, output, cache writes, cache reads), the amount billed to your credit balance (zero on an own key), and — for managed calls — an internal identifier of the Tejma key that served it, which only Tejma's operators can see. These records contain token counts and amounts, never message content.

Billing information. Each seat's plan history and billing start, task-execution counts per cycle, the organization's prepaid credit balance and its full ledger (purchases and the bonus credit each earned with the rate applied, gifts, grants, per-call AI debits, overage debits, corrections), the overage opt-in, per-call credit holds in flight, and Stripe identifiers (customer, subscription, and checkout-session ids). Card numbers are entered on Stripe's hosted pages and never touch our systems; the billing address and optional tax ID you enter at checkout are stored by Stripe on your customer record.

Logs and diagnostics. Our servers write operational logs, retained by our hosting provider for a limited period. Beyond request metadata, timestamps, and error traces, these logs include: one line per AI call with the task identifier, agent name, attempt and turn numbers, the names of the tools the agent called with whether each succeeded and, on failure, the first part of the tool's error text, and token counts (never the model's prompt, its reply, or tool inputs and results); the addresses messages were delivered to; the reasons judges gave; the names of attachments as they are stored; the addresses of connected mailboxes; the identifier and display name of a member being deleted; and the full path of each API request, including any search text or sign-in parameters it carried. Webhook delivery records keep ingestion and payment processing reliable.

3. How We Use Information

We use the information above to run the service you configured: to ingest and store messages, extract the tasks they contain, route each task to your agents, evaluate your approval workflows, execute approved tasks, send the outbound messages your agents are directed to send, tell Tejma's connector services whether a member may use them, meter task executions against your plan and AI calls against your credit balance, show you what your usage costs, collect payment, prevent fraud and abuse, secure the service, keep the consent records the law expects, and provide support. We also use aggregate, non-content operational data to maintain reliability and to manage the capacity of our managed AI key pool.

We do not sell your data, use it for advertising, or use your content to train AI models.

Message content is processed only to provide the service to you, by us and by the service providers listed below.

4. AI Processing

Content is sent to Anthropic. To extract tasks from a message, route a task to an agent, evaluate an approval workflow's judge, or execute a task, the relevant content is sent to Anthropic's Claude API: the message text and its thread transcript, task descriptions and the completion notes of prerequisite tasks, your skill instructions, the list of a message's attachments and the content of any attachment the model chooses to read (within size limits — currently 16 MB for a PDF, 5 MB for an image, 1 MB for a text file, and 20 MB per run), and the results of the tools the agent calls. We use Anthropic's prompt caching, so a recent prompt prefix may be held by Anthropic for a few minutes to reduce cost.

Whose Anthropic account. Under managed AI billing (the default), calls are made with API keys that Tejma holds, under Tejma's commercial agreement with Anthropic: Anthropic is our service provider and subprocessor for that processing, and your content is handled under Anthropic's commercial terms as they apply to Tejma. Under own-key billing, calls are made with the key your organization supplied, under your organization's own agreement with Anthropic, and Anthropic bills your organization directly. In both modes, Anthropic's commercial terms provide that API inputs and outputs are not used to train Anthropic's models absent an opt-in, and Anthropic retains and secures API data as described in its commercial terms and privacy documentation.

Connectors. When an agent executes with connectors, its tool calls run server-side at Anthropic's API: the connector's URL and your authorization token for it are passed to Anthropic for the duration of the call, refreshed on each turn, and Anthropic's servers call the connector with that token, so task-related content may be sent to those connector servers and their replies pass back through Anthropic. The service connects to a connector directly only to list its tools. For connectors you register, you choose those servers; review their privacy practices before wiring them to an agent. A Tejma connector additionally verifies your entitlement with the service: it sends us your user identifier and receives back whether your seat holds a current, unparked install and grant, your organization identifier, the install's identifier, and your granted roles — never your name or email address.

Metering. Every call produces an AI usage record (above) used for billing and reporting. Your organization's billing-role holders can review usage by job, model, user, agent, and day, each task's AI cost, and the ledger entry each managed call produced. Under managed AI billing the amount charged is the per-model price in Tejma's catalog, which Tejma sets and which may differ from Anthropic's own prices; the current prices are shown to billing-role holders in the dashboard's pricing page. Customers never see which Tejma key served a call.

5. Service Providers and Recipients

We share information only with the providers needed to run the service, each bound to use it solely to provide their service to us and to you:

  • WorkOS — authentication, sessions, organization and member management, invitation emails (identity data).
  • Anthropic — AI processing of message and task content: under Tejma's own Anthropic account for managed AI billing, under your organization's own account when it supplies its key.
  • Nylas — email connectivity: mailbox OAuth, message sync, thread listing, attachment retrieval, and sending (US region by default; EU region available).
  • Slack and Discord — receiving messages from, and sending messages to, the channels you connect, through their APIs.
  • Stripe — payment processing: checkout, recurring subscription billing, invoices, customer portal, and dunning (billing contact, billing address, tax ID, and payment data).
  • Railway — cloud hosting: runs our servers and the Postgres database that holds messages, tasks, usage records, and account data, and retains our server logs.
  • Cloudflare — R2 object storage holding message documents and attachments, and the network tunnel that delivers inbound webhooks to our servers.

Tejma's connector services. When you use a Tejma connector, that service receives your user identifier, organization identifier, install identifier, and granted roles from the dispatcher (above), and receives whatever task content the agent passes to its tools, through Anthropic. Each Tejma connector's own privacy policy describes what it does with that data.

Recipients of the messages your agents send. Outbound messages and completion reports are disclosures of your content to their recipients, which your configuration determines: fixed recipients you set, the participants of the thread a message answers (who may be outside your organization), or, where you allow it, up to 50 addresses the agent chooses itself. A completion report includes a short description of the task and a link to it in the dashboard, which opens only for a signed-in member who owns it. Delivery is at least once, so a recipient may occasionally receive a message twice.

Beyond these, we disclose information only if required by law or legal process (we will notify the organization's administrators unless legally barred), to protect the rights, safety, or security of Tejma, our customers, or the public, or as part of a merger or acquisition — in which case this policy continues to apply until you are notified otherwise. A current subprocessor list for business customers is maintained in the Data Processing Agreement.

6. Data Retention and Deletion

Content is kept until it is deleted. Messages, their thread transcripts, their stored documents, and the audit trail of messages sent are retained so your tasks keep working even after the originating email, chat message, or channel connection is gone — disconnecting a channel stops new ingestion but does not delete what was already stored. The service applies no time-based deletion of its own: content remains until it is deleted by one of the controls below or at your request.

Deletion controls in the service. Deleting a stored document is immediate and permanent: its bytes are removed from the object store before its record. Deleting a task — one at a time, or several waiting tasks at once — marks it deleted and hides it, but its record, its message, and any evaluations and usage records about it remain; tasks that are running or completed cannot be deleted, and archiving a finished task only hides it from the board. Agents can be deleted once all their tasks are finished; finished tasks keep the agent's name. Deleting a connector or workflow removes it, and disconnecting a channel or a channel account stops ingestion and revokes the connection at Slack, Discord, or your email provider. Messages, their thread transcripts, and the send history cannot be deleted individually in the service.

Deleting a seat. A users-role holder who deletes a member purges that seat in one step — messages, thread transcripts, documents and their stored files, tasks, agents, approval workflows, connectors and their stored tokens, connector grants, channels, send history, roles, plan timeline, and usage counters — revokes the member's sessions, and revokes the seat's channel connections at their providers. Connector authorizations are not revoked at the accounts that issued them; revoke them there. To erase particular messages without deleting the seat, or to delete an organization's content entirely, a billing- or users-role holder emails us at the address below and we perform the deletion; cancelling a subscription deletes nothing by itself.

What outlives a seat. Terms acceptance records are kept as evidence of consent. AI usage records and credit-ledger entries are kept as billing and audit records, with their references to the deleted user, tasks, and messages cleared; they hold token counts and amounts, never message content. The email address on an invitation record, the identifier of a deleted member recorded as the grantor of other members' roles and connectors, Stripe's records of the organization, and server logs remain until they age out on their own schedules.

After an organization closes. When an organization asks us to close, we delete each of its seats as described above. Billing records, ledgers, and usage records are retained as long as tax and accounting law requires. Encrypted credentials are deleted with the records that hold them; database backups and server logs are retained by our hosting providers for a limited period and are protected until they age out.

7. Security

We protect information with measures matched to what the service holds:

  • stored secrets — channel and connector tokens, customers' own Anthropic API keys, and Tejma's managed Anthropic API keys — encrypted by the application at rest with AES-256-GCM and treated as write-only; message content, documents, and other records rely on the at-rest encryption of our hosting and storage providers;
  • strict per-tenant isolation: every record belongs to one user, and another user's identifiers behave as if they do not exist;
  • signature verification on every inbound webhook (Nylas, Slack, and Stripe events are each cryptographically verified before processing);
  • OAuth 2.0 bearer-token authentication with JWT signature and issuer verification against WorkOS on every API and MCP request, and TLS in transit;
  • role-gated controls: billing, credit, and credential changes require the billing role; member, invitation, grant, and role changes require the users role; and a Tejma connector learns only a member's identifiers and roles, never their name or email;
  • a managed key that Anthropic rejects is deactivated automatically, and the key that served a call is recorded for Tejma's audit only.

Tejma personnel access. Tejma's operators sign in through a separate identity tenant that no organization member can reach. Its controls cover plans and prices, every organization's billing standing, seats, usage counters, and credit ledger, credit adjustments, cross-organization AI usage records, the managed key pool, agent and connector templates, member lookup by email address, member deletion, and role changes; none of them displays message content. Operators also have a maintenance tool that can re-read a connected mailbox's messages from our email provider for a time window, to recover mail a notification missed; it writes the recovered messages' subjects to our logs. Personnel with access to our infrastructure can reach stored content, which is not encrypted by the application; they are bound by confidentiality and access it only to operate, secure, and support the service.

No service can promise perfect security. If a breach affects your personal data, we will notify affected organizations' administrators without undue delay and as the law requires.

8. Your Rights and Choices

Access, correction, deletion, portability. You can read your content in the service at any time, and delete documents and, through a users-role holder, whole seats. For anything else — a copy of your data, correction, deletion of particular messages, or deletion of an organization — email us at the address below and we will respond within the time the applicable law sets. If your personal data reached us through a customer's connected channel (for example, you emailed one of our customers), we will refer your request to that customer, who controls that content, and assist them in honoring it.

Revoking connector and channel access. Disconnecting a channel in the service revokes its connection at Slack, Discord, or your email provider. Deleting a connector deletes our copy of its token but does not revoke the authorization you granted; revoke it in the account that issued it — your Tejma account's connected applications for a Tejma connector, or the third-party server's own account settings otherwise.

Billing mode is a choice. Your organization's billing-role holders decide whether AI calls run under Tejma's managed account or under your organization's own Anthropic account, and can switch at any time; the switch applies from the next call.

Region-specific rights. Depending on where you live, you may have statutory rights — under the GDPR or UK GDPR (access, rectification, erasure, restriction, portability, objection, and complaint to a supervisory authority) or the California Consumer Privacy Act (to know, delete, correct, and not be discriminated against; we do not sell or share personal information as the CCPA defines those terms). We honor these rights regardless of which law applies.

Communications. Notices about failed payments come from Stripe on our behalf. The service itself does not send alerts about your credit balance or price changes; your billing standing and the current prices are always visible in the dashboard. We do not send marketing communications through the service's data.

9. Email Provider Account Data

What we ask for. Connecting a mailbox for incoming mail requests read access only: for Google, the openid, email, and gmail.readonly scopes; for Microsoft, openid, offline_access, User.Read, and Mail.Read. Connecting a mailbox for sending additionally requests gmail.compose for Google, or Mail.Send and Mail.ReadWrite for Microsoft. A connection that grants only part of what was requested is refused, and we record the scopes actually granted to decide what the connection may do. Read access is used to receive new mail and, for a reply, to read the earlier messages in its thread — which may include your own sent replies and messages from before the connection; emails too large for a notification are read in full from the provider. We do not store the raw notifications the provider sends us, only the message as described above. If the same mailbox is connected again by another member of your organization, its future mail goes to that member.

Google. If you connect a Gmail or Google Workspace mailbox, our use of information received from Google APIs adheres to the Google API Services User Data Policy, including its Limited Use requirements: Google user data is used only to provide the message-ingestion, task, and reply features you see, is never used for advertising, and is never transferred except to provide those features (including the AI processing described above), for security, or to comply with law.

Microsoft. Microsoft 365 and Outlook mailboxes connected through the service are handled the same way, in line with Microsoft's API terms.

10. Children, International Transfers, and Changes

Children. The service is a business tool and is not directed to children under 16; we do not knowingly collect their data.

International transfers. We operate from the United States and our providers process data primarily in the United States (email connectivity can be pinned to the EU region). Where the GDPR applies to transfers, we rely on appropriate safeguards such as Standard Contractual Clauses, as detailed in the Data Processing Agreement.

Changes to this policy. We will post updates here and change the date at the top. When we publish a new version, the service asks you to accept it before you continue.

Contact. Tejma LLC · Reno, Nevada · ai.assistants@tejmallc.com · tejmallc.com

Matej Meciar · Tejma LLC

925-963-4257 · ai.assistants@tejmallc.com · tejmallc.com

Tejma LLC · Reno, Nevada · Practical workflow automation for business teams. · 925-963-4257 · ai.assistants@tejmallc.com

Tejma Dispatcher: Terms of Service · Privacy Policy · Data Processing Agreement

Tejma Books: Terms of Service · Privacy Policy

Tejma Work Order Scheduling: Terms of Service · Privacy Policy