Frequently asked questions

Describe the job once. ChatFuse puts together an agent that hooks into your apps, picks the right model for each part, and checks with you before it does anything. It lives in one workspace with your chats, Canvas and Library, and you're never tied to one AI provider.

ChatFuse is model agnostic. We support over 100 top models, including the latest from OpenAI, Anthropic, Google, Meta, and more, and we add new ones as they come out.

ChatFuse uses the Orchestrator to automatically route each request to the model best suited for it. The Orchestrator analyzes your query in real time and selects the model that delivers the best output. You never have to choose manually.

The Orchestrator is the core of ChatFuse. It automatically classifies your request and decides which AI model to use based on the task, such as coding, research, writing, or image generation. This ensures every response comes from the model most capable of handling that type of query, providing higher accuracy and better results without any extra setup.

We prioritize improvements using product reliability, customer feedback, and the workflows people use most. New capabilities are released gradually after they meet our quality and privacy standards.

Extensions are the Apps and Skills you can use with ChatFuse. Apps connect services where your work already lives, while Skills provide reusable instructions for completing repeatable tasks. Extensions are currently available to a limited private preview audience.

Extensions are in a limited private preview, and ChatFuse controls which accounts are enabled. A workspace administrator cannot turn them on for you. Chat and Tools keep working normally in the meantime. Contact ChatFuse support if you believe your account should already have access.

An App gives ChatFuse permission to access or act in an external service you connect. A Skill describes how ChatFuse should complete a repeatable workflow. A Skill may recommend or require one or more Apps, but installing a Skill does not grant access to an external account.

A Persona controls who ChatFuse is, including its role, expertise, and communication style. A Skill controls how a specific task should be completed. You can use a Persona and a Skill together.

The signed in Extensions directory is the authority. ChatFuse dynamically lists every App available through ChatFuse's managed connector layer with supported OAuth or API key authentication and a genuine connection configuration. Apps with reviewed actions additionally expose bounded safe read or confirmation required non destructive execution. Unsupported auth, stale, malformed, provider disabled, unconfigured OAuth, and changed contract Apps stay hidden or unavailable. Google Drive, Gmail, and Notion retain narrower provider specific handling where required.

Sign in, open Extensions, select Apps, and choose Connect on a supported App. Most Apps send you to the provider's own permission screen, which you complete yourself before returning to ChatFuse. Apps that authenticate with an API key instead open a short form in ChatFuse where you enter the credentials the provider issued to you. ChatFuse passes those credentials to the connector layer and does not store them. Either way, a genuinely active connection appears in Installed and can become available in the chat Extensions picker.

Open the plus menu beside the chat composer, choose Extensions, and select one or more connected Apps plus an optional installed Skill. Choose the exact account when an App has multiple connections. App logos and chips appear at the composer, and the server authoritative selections remain scoped to that chat until you remove them.

The ready made catalog contains 48 production qualified built in Skills across 12 categories. Private and user created Skills are separate and do not count toward that total. The live Skills directory and Chat's catalog answer show current names, categories, install state, and App dependencies.

In the Extensions chat preview, you review and confirm each supported write before it happens. Destructive actions aren't available there. Agents have separate delivery controls: an exact approval or a matching Standing delivery permission you've approved. The help bot can't confirm or send on your behalf.

Access depends on the App and the permissions shown during connection. ChatFuse exposes only the tools it has enabled for that App. Review the provider permission screen before connecting and disconnect the App whenever you no longer want it available.

Open Extensions, select the App in Installed, and choose Disconnect. ChatFuse immediately stops using that connection. Where supported, the external authorization is revoked as part of disconnection; you can also review connected app access in the external service.

No blanket access should be assumed. Access is limited by the permissions you approve and the tools ChatFuse enables for that App. Review the permission screen before connecting and disconnect the App whenever you no longer want it available.

In Auto mode, ChatFuse can select an eligible model based on the task and the capabilities required. If a manually selected model cannot use Apps, ChatFuse asks you to switch to Auto or remove the App.

Secure account connection and reviewed ChatFuse actions are separate capabilities. The catalog can connect many services, but only Apps with explicitly enabled and tested tools can be used for in chat reads or writes.

Open Extensions and check the App's connection status. Reconnect it if permission expired or was revoked. If the issue continues, retry later or contact support with the App name, time, and visible error wording. Never send support a password, OAuth code, access token, refresh token, or provider API key.

Save a repeatable job as steps you can review and run again. Agents can use selected sources and available Apps, process information, save a result and deliver it through supported destinations. The editor shows which capabilities are ready and which need setup.

A Persona supplies reusable instructions for a conversation. A Skill describes how to do a task. An Agent saves the job with its steps, sources, published version and run history. Using a Skill doesn't give an Agent extra account permissions.

Describe the job in Chat, then review the proposal's steps, sources, accounts and estimate. Save and check the draft. You can inspect it visually or use Ordered steps on mobile. Publish the version you want to run. Later draft edits don't change work already started.

No. A sample can't send messages, write to an external App or activate a trigger. It may use paid reads within the allowance you review first. Mock content is identified, and a successful sample doesn't prove that live delivery is available.

Yes, with an activated schedule or supported event trigger and the required background permission. It still needs current access and available credits. Revoking a connection blocks affected work. For schedules, confirm the timezone and preview the upcoming starts; a delivery deadline doesn't guarantee arrival.

Each external action needs approval for the exact reviewed content and destination, or a matching Standing delivery permission you've approved. That permission has specific boundaries and limits. Connecting an App or activating a schedule doesn't authorize every future send.

You can receive an optional SMS notice with a link to your result when the recipient, consent and sender are ready. Verify the number and save SMS permission in the agent's settings. The link requires access to the result. Private result content isn't included in the text. Reply STOP to opt out; START requires fresh consent in ChatFuse.

Paid work reserves credits before it starts and settles against recorded usage. Run and account limits apply. A budget caps spending but doesn't guarantee completion, and background work doesn't automatically buy more credits. Reusing authorized results can avoid repeated work; delivery may still cost credits.

Pause stops future automatic starts. A paused Agent can still run manually. Cancel run requests that remaining work stop, but a message already submitted may still arrive. Usage already incurred still needs to settle. History shows the result and delivery states separately.

Open History and read the reason. The run may need approval, repaired access, an available capability or a budget change. Reconnecting an App alone doesn't restart it. Review run permissions and use Resume run separately when offered. If a send is uncertain, check its status before trying again.

Access checks apply to your selected sources, agent and results. A result link doesn't make private content public. A review invitation shares the material you've chosen without giving the reviewer your App permissions or sending authority. Saving content from a conversation with restricted storage requires a separate disclosure and acknowledgement.

ChatFuse uses encrypted transport (TLS), encrypted storage, strict access controls, and audit logging. We apply least privilege access to internal systems.

No. Your conversations are not used to train AI models.

Your data is protected by business grade security controls on Google Cloud infrastructure, including encryption in transit and at rest, role based access controls, and continuous monitoring.

All ChatFuse data is securely stored on Google Cloud servers located in the United States. We comply with global data privacy laws including GDPR and CCPA to ensure regional and international data protection standards are met.

Yes. Export your account data from Profile settings before deleting the account. Deletion signs you out and schedules permanent removal after a 30 day recovery window. The flow also requests cancellation of an active subscription and sends a restore link. During the recovery window, choose Restore from that link or from the recovery screen shown when you sign in again. Signing in alone does not restore the account.

ChatFuse employs multiple layers of cybersecurity protection: AES-256 encryption in transit and at rest, regular penetration testing, automated threat detection, secure API gateways, and isolated processing environments. Our infrastructure is monitored 24/7 for suspicious activity and potential breaches.

We use layered security controls including encryption, access management, network protections, and monitoring. No system is risk free, but we design for rapid detection and response.

ChatFuse is built on compliance ready infrastructure and supports GDPR/CCPA aligned processes. For regulated use cases, we provide legal/compliance documentation and contractual controls.

We review valid legal requests and respond according to applicable law. We limit disclosure to what is legally required and maintain strict internal handling procedures.

You can upgrade from a monthly to a yearly plan anytime. The yearly plan starts immediately and replaces your monthly billing cycle.

ChatFuse uses credits to measure usage. Tasks such as image generation and writing use credits based on the work involved. Pro Monthly includes 20 million credits per paid month. Pro Yearly grants its allowance for the paid year. Renewal follows the confirmed subscription period. Trial credits are separate and last for the trial period.

Yes. Pro Monthly includes 20 million credits per paid month. Pro Yearly uses an annual allowance. Usage depends on the tasks you run. The app notifies you as credits run low, and eligible accounts can buy additional credits.

Yes. You can cancel whenever you like, and your plan will remain active until the end of your billing period.

Our support is fast, friendly, and available to everyone. You can use the ChatFuse Assistant to find answers or get help exploring the site. If you need direct assistance, you can always reach a human. We use HappyFox to organize and track all requests so nothing is missed.

Yes. ChatFuse supports business use with Business Associate Agreements for healthcare organizations and Data Processing Addendums for business customers.

Yes. Business clients can access our API for direct model integration, workflow automation, and large scale orchestration. We also support custom setups for teams that require secure, private deployments of ChatFuse technology.