Frequently asked questions

Thor is an AI teammate and context layer for software teams. It's context infrastructure for AI-native teams: one living record of what your team decided, kept current from Slack, meetings, GitHub, Linear, and your docs, so the next person or coding agent never starts from a guess. Thor drafts the updates that keep that record honest, and a person approves them before anything official changes.

Thor is built by the team at Zenhub. You can reach the team at support@thor.ai.

Thor is purpose-built for software engineering and technical teams that are moving too fast and feeling overwhelmed with administrative overhead. It's ideal for team leads, engineering managers, and founders who feel the pain of coordination tax most acutely.

Thor proposes and a person approves. Thor drafts the change, a new issue, a new owner, a status update, a doc edit, shows you where it came from, and waits for someone to accept it before anything in your tools changes. Nothing becomes official just because an AI sounded sure.

No. Thor works inside the tools your team already uses, and it won't ask you to move your work somewhere new. Slack stays your chat, Linear or GitHub stays your tracker, Notion or Google Docs stays your docs. What Thor adds is the upkeep. When a meeting or a Slack thread changes the plan, Thor drafts the update to the ticket or the doc, someone approves it, and the tools go back to telling the same story.

Thor doesn't wait to be asked. A general assistant answers from whatever you paste into the chat, so two people asking the same question can get two different answers, and nothing gets written down as decided. Thor reads the work itself, keeps one shared record of what's decided and what's gone stale, and comes to you when a commitment is aging or a ticket no longer matches what the team agreed. Every answer shows where it came from. General AI assistants are designed to answer general questions, from medical diagnosis to child care to weather reports. Thor is focused on a narrow mission -- keeping your team's work honest and up to date.

Nope. You have enough bots to deal with. Thor isn't an awkward guest you have to remember to add to every call. It works by hooking into the meeting transcripts your team already generates in Google Meet, Zoom, and Microsoft Teams.

Not at all. You can initiate the setup in minutes using standard, secure third-party authorization like GitHub OAuth. If an integration requires broader admin approval, Thor will provide a simple, shareable link for your workspace admin to grant the necessary permissions.

We are currently in an early access program, working closely with our design partners to finalize a fair and flexible pricing model. For now, early access is provided at no cost. Request access today to lock in preferential pricing when we launch.

No. Search helps you find what was written down. Thor keeps track of what's happening around the work: what changed, what's blocked, who owns what, how one work item relates to another, and where attention is needed. Thor also proactively reaches out when it detects drift from the plan or missed commitments. Thor is more like a team member that you hired. Search is a part of it, but Thor is more than that.

No. Agents need operational context to act well. Humans need it to know where to pay attention. Thor gives both a clearer picture of what's actually happening.

Thor connects to the services where the work already happens: Slack or Microsoft Teams for chat, GitHub and Linear for issues, Google Meet and Zoom for meeting transcripts, and Notion and Google Docs for documents. Coding agents read the same checked context through Thor MCP. Chat plus a tracker is enough to be useful after your next standup.

After your meeting ends, Thor reads the transcript and figures out what actually needs to happen next. Then Thor drafts a plan: either the action items can be converted to issues, or a document update can be triggered to capture any decisions that were made, or Thor can follow up with Sarah about that API thing you discussed on the call but already forgot about. You look over Thor's suggested next steps, give it the thumbs up, and Thor actions it. Nothing lands in your tracker until someone says yes. Transcription needs to be switched on in Google Meet or Zoom, since Thor never joins the call.

Thor MCP is how AI assistants and coding agents read Thor's context. MCP, the Model Context Protocol, is the standard way an AI tool connects to an outside source, so Cursor, Claude Code, and other agents can ask Thor what's happening on a project, what was decided, and what's gone stale, instead of rebuilding that picture from raw Linear and GitHub calls on every prompt. Agents read the same checked record your team sees, so the human and the agent aren't working from different stories.

Raw MCP connections give an agent access to your tools. They don't give it an understanding of how the work fits together, so every question turns into a fresh reconstruction job. Wiring it up yourself is also ongoing upkeep, since OAuth tokens expire and someone has to keep them alive. Thor connects to Slack, GitHub, and Linear natively at the API level and keeps one checked record that people and agents both read. We still use MCP for the few things we can't reach natively, and we keep that part as small as we can.

No. Thor doesn't write code, and it won't replace Cursor, Claude Code, Factory, Devin or whatever your engineers run. It feeds them. Thor handles the coordination around the code: turning a meeting into tracked issues, keeping a ticket honest when the plan changes, noticing a commitment that's gone quiet, and handing a coding agent a spec that matches what the team actually decided. Thor is the agent that gets your coding agents the execution context and plan to make them more effective.

Those tools are strong inside their own walls, and Thor doesn't try to beat them there. The gap is between tools. A meeting changes the plan, the decision lands in Slack, the ticket still says the old thing, and the doc says a third thing. Thor's job is keeping one honest story across all of them. The Linear AI doesn't know what's happening in Slack or Notion, but Thor does.

Thor is specifically tuned for software development and product team conversations. However, our key feature is the "human-in-the-loop" design. Thor always proposes a plan for your approval before taking any action, giving you the chance to add any necessary nuance or corrections.

No, the opposite. By centralizing follow-ups and creating structured action plans, Thor reduces the random "just checking in" pings on Slack. You can configure how and when Thor communicates to fit your team's existing workflow and minimize unnecessary noise.

We are constantly expanding our ecosystem of connected tools based on customer feedback. If you have a specific tool you'd like us to support, we'd love to hear about your use case. Please reach out to our team at support@thor.ai.

Yes, Thor is fully SOC 2 Type 2 compliant. Thor is a new product from the team behind Zenhub, who have a decade-long track record of building enterprise-grade developer tools with security at their core. We apply the same rigorous security and compliance standards to everything we build.

Absolutely not. Your data is yours alone. We never use customer data, conversations, or meeting transcripts to train our models.

Thor is designed to be policy-aware. It inherits the exact same permissions as the user who connects it. If you don't have access to a specific repository or document, neither does Thor. It can only see what you can see and only take action where you have the appropriate rights to do so.

No. Thor's integration with tools like GitHub is strictly limited to project management functions like creating, updating, and commenting on issues. It has read access to your codebase in order to understand the context of the work and access the relevant documentation (typically stored in markdown files). Thor cannot push directly to your codebase.

Yes. Thor maintains an internal audit trail of every action it takes on your behalf. At the moment these logs aren't directly exposed through the product, but can be provided upon request.