Frequently asked questions.

Zihin connects AI agents to company systems and processes, with permissions, approvals and monitoring defined for each operation.

The Zihin REST API has its own request contract. Integrate through HTTP or use the documented packages. Existing integrations may need authentication, payload and response-handling changes.

You can select models from the catalog or use automatic routing. Switching depends on each model’s capabilities and accepted parameters; configure fallback for your use case.

Integrations use documented HTTP and MCP interfaces. Changing platforms requires reviewing contracts, configurations, data and tools; automatic migration without changes is not guaranteed.

The Trial lets you evaluate the platform for 15 days. Check the plans page and sign-up flow for own-key requirements, limits and terms applicable to your account.

No. The Trial doesn't ask for a card. You only provide a payment method when you decide to subscribe to a paid plan.

Trial, Core, Pro and Business serve different usage and implementation needs. The plans page lists published prices and features; specific terms are confirmed when subscribing.

The platform subscription and model usage are separate components. With BYOK, your provider bills usage on your key. Check allowances, managed pool access and terms in your subscription.

Yes. You can upgrade or downgrade whenever you want; the change takes effect at the next cycle.

Continuation depends on the limit reached and the account configuration. Check usage in the console and available expansion options; using your own key does not remove platform limits.

Add-ons expand the subscription, for example with usage capacity or seats. Check the availability, price and validity of each option when subscribing.

Yes. You keep access until the end of the paid cycle, and self-service plans have no lock-in.

See the partner program. For nonprofits and educational projects, contact the team about available terms.

You connect your own AI provider key, and Zihin runs the agents using that key — with no markup on tokens. You pay the provider directly for consumption and Zihin for the platform. BYOK is available on every plan and required on the Trial.

Models from Anthropic (Claude), OpenAI, Google (Gemini), and xAI (Grok). You can set a primary model and a fallback chain — if the primary becomes unavailable, Zihin automatically switches to the next in line.

Own-key and managed-pool availability depends on your plan and account configuration. Check the requirements before starting a trial or subscribing.

No. Zihin doesn't train models on your data. On BYOK, calls go through your own provider account, subject to its terms (the commercial APIs from Anthropic and OpenAI, for example, don't use API data for training).

Editor creates and edits agents (including flows, connections, and configuration). Member uses the finished agents (chat, WhatsApp, Slack). Both count as plan users.

Admins manage the account, users and settings according to their permissions. Seat counts and billing access follow the plan and assigned role.

Check included seats and expansion options on the plans page and when subscribing. Agent access also depends on each person’s role and permissions.

Four-level RBAC: owner → admin → editor → member, with permissions enforced in both the interface and the API.

Agents use configured API tools and MCP servers. For ERP and database data, Zihin Replay keeps a read-only copy in your network and MCP Bridge provides read-only queries over contracts on that copy; the agent has no direct database access. Each integration requires compatibility, authentication and scope checks.

An agent can call the tools you make available under configured permissions and policies. An integration may be read-only, like MCP Bridge, or allow specific actions; write access is not automatic.

Your team defines configurations and instructions; memory can store and retrieve context across sessions. This is not automatic training and does not guarantee identical responses. Review results and changes according to your process.

You can provide information through tools, files and integrations, and configure process steps and human decisions. Sources, data refresh, permissions and coverage must be defined for each implementation.

Zihin Chat provides access to agents enabled for the team. Other channels can be configured through integrations, such as WhatsApp via Twilio and Slack or Teams via webhook/API. Requirements and availability vary by channel and account.

Configured schedules can start agents at defined times. Check time zone, input data, result destination and feature availability on the account.

Zihin Watch combines operational observability and email reports. Weekly summaries and monthly reports cover executions, AI costs and issues needing attention for designated recipients and accounts with delivery enabled. Periodic reports complement execution monitoring; access and retention depend on configuration.

Yes, MCP-compatible clients can connect to the Zihin server. Available tools depend on role and permissions; actions require approval when configured policies call for it.

Configured analyses can generate tables, files and reports accessible to the team. Formats depend on available tools. Results can include the coverage of the queried data.

Implementation defines steps, owners and points requiring human decisions. The team can review context and record a decision before the planned continuation. This depends on the configured process and features enabled on the account.

Through contextual security policies (CSP): you define what the agent must never expose (tokens, credentials, sensitive fields) and those rules are enforced on every call. Add to that RBAC, an encrypted secrets vault (AES-256), sensitive data masking, and an audit trail.

The platform provides features that support compliance — sensitive field masking, access control, encryption, and an audit trail — and we provide a DPA (Data Processing Agreement) as an annex to the Terms of Use, covering Controller and Processor roles under LGPD and GDPR.

Yes. The platform is multi-tenant with per-tenant isolation at both the database and application levels.

Credentials are stored in an encrypted vault and resolved by authorized integrations. Access control and care with data sent to tools remain part of configuring the operation.

Discuss the identity, SSO, audit and retention requirements for your operation with the team. Confirm the available scope in the proposal and contract.

The site does not present SOC 2 or ISO 27001 certification. Ask the team for available security evidence for your assessment; technical controls and a DPA are not equivalent to certification.

Report vulnerabilities to contact@zihin.ai with enough context to investigate, without sending credentials or sensitive data. Applicable data-incident procedures are described in contractual and privacy documents.

The platform runs in the cloud. Replay, MCP Bridge and Tunnel let you connect data and services in the company’s infrastructure, depending on the deployment. This does not mean the whole platform is installed on premises.

Availability and implementation scope must be confirmed with the commercial team based on infrastructure, data contracts and access requirements.

See the Privacy Policy and DPA for applicable hosting, providers and transfers. Data residency requirements should be assessed for your scenario when contracting.

Yes. You can request export and deletion of your data via contact@zihin.ai — we respond within 30 days, as per the Privacy Policy.

Trial: community support. Core and Pro: standard support. Business: dedicated support with an SLA.

Availability targets and support terms follow the Terms of Use and, where applicable, the specific contract. Refer to those documents for scope, criteria and conditions.

Discuss volume, integrations, identity and contractual requirements with the team. The proposal should define scope and terms before implementation.

A continuity option after going to production: instead of handing operations to your internal team, Zihin keeps managing and evolving your agents.

Start by describing the process, systems involved and expected result. Discovery guides planning, piloting and implementation, with defined owners, access and monitoring.