Frequently Asked Questions

Real questions, real answers. No marketing fluff.

What is Confab?

Confab is a team AI collaboration tool. Your team talks to AI from Slack, Discord, or the web, and the AI has full context on your codebase. You connect your own API keys for Claude, ChatGPT, Codex, or Gemini. The AI can answer questions, write specs, make code changes, and open PRs. Everything happens in the chat tools your team already uses.

How is Confab different from ChatGPT or Claude?

ChatGPT and Claude are general-purpose AI assistants. Confab connects those same models to your codebase and your team's chat channels. When your PM asks a question in Slack, the AI answers with knowledge of your actual code, not generic advice. When your dev asks for a code change, the AI commits to a branch and opens a PR. The AI providers give you the brain. Confab gives your team a shared way to use it.

How is Confab different from Viktor?

Viktor is a task execution bot where each person DMs the bot individually. There are no shared threads or team collaboration. Confab is built for teams from the ground up: multiple people work with AI in the same Slack channel with shared project context, role-based permissions, and an audit trail. Viktor also uses opaque credit-based pricing ($300-400/mo in real-world use). Confab charges a flat team price starting at $49/mo for 5 users.

How is Confab different from OpenClaw or hosted OpenClaw services?

OpenClaw is a single-user personal AI agent. It has no team collaboration, no permissions, and no multi-user support. Hosted OpenClaw services charge per instance, so a 5-person team pays for 5 separate containers ($95-395/mo). Confab is multi-tenant with role-based permissions: your PM gets read-only access, your dev gets code editing, and your admin controls who can do what. One flat price covers the whole team.

Can I compare responses from multiple AI models?

Yes. Confab's /compare command sends the same question to multiple AI models simultaneously and shows the responses side by side in your Slack or Discord thread. You can compare Claude, ChatGPT, Codex, and Gemini on the same question. This is useful for architecture decisions, code review, and estimating confidence. If all models agree, you can trust the answer more. If they diverge, the question needs more thought.

What AI models does Confab support?

Confab supports any AI model you can connect via API key: Claude (Anthropic), ChatGPT and Codex (OpenAI), and Gemini (Google). You bring your own API keys and pay your provider directly at standard rates. Confab never marks up your AI costs. The Personal plan supports comparing 2 models at a time. Pro and Enterprise support comparing all connected models.

Does Confab work with Slack?

Yes, Slack is a primary channel. You add Confab to your Slack workspace, bind it to a project, and your team can DM the bot or interact with it in channels. Each thread gets its own AI session with isolated context. The AI knows who is asking and respects their permission level. PMs can ask questions, devs can request code changes, and the whole team sees the work happen in the open.

Does Confab work with Discord?

Yes. Confab supports Discord servers with the same features as Slack: channel-level AI, thread-based sessions, and role-based permissions. Many developer teams and open-source communities use Discord, so it's a launch channel alongside Slack.

Can non-developers use Confab?

Yes, and this is one of Confab's main design goals. Product managers, founders, designers, QA engineers, and compliance teams can all interact with the AI from Slack without touching a terminal or knowing git. Permissions control what each person can do: a PM can ask questions and edit documents but can't trigger code deployments. A founder can describe a new idea and the AI will create a repo and starter doc automatically.

How does the auto-repo creation work?

You describe a new idea in chat. Confab confirms your description, suggests a repo name, and when you approve, creates the repo on your connected GitHub, GitLab, Bitbucket, or Forgejo account. It commits a starter document synthesized from your description and sends you a link. The whole process takes about 30 seconds. No git knowledge required. This is useful for PMs and founders who want to start a project without involving a developer.

Does the AI write directly to my main branch?

No, never. Every code change, spec, and document goes to a working branch. Code repos get a pull request for human review. Doc-only repos can enable auto-merge if you prefer. Either way, there is a full audit trail of who asked for what, what the AI did, and what changed. The AI never writes directly to main. This is an architectural guarantee, not a setting.

What git providers does Confab support?

Confab supports GitHub, GitLab, Bitbucket, Forgejo, and Azure DevOps. You connect via OAuth during project setup. Confab auto-registers webhooks so your repo context stays fresh (near-instant sync on push, not polling).

How does Confab handle permissions?

Admins configure two permission policies: one for users who DM the bot directly, and one for users in channels. There are five permission categories: Answer Questions, Multi-model Questions, Edit Documents, Edit Code, and Trigger Workflows. Each permission can be set to Allowed, Denied, or Confirm via DM (which escalates to an admin for approval). This means a PM can ask questions but not trigger code changes, unless an admin approves it.

Is Confab secure?

Security is a core architectural requirement, not a bolt-on. API keys are encrypted at rest with per-tenant keys. Repo data is isolated per tenant in ephemeral containers. AI outputs are scanned for secrets, malicious patterns, and suspicious dependencies before any commit. Agents cannot access .env files or credentials. Every AI-generated change goes through a branch and PR. Confab uses standard BYOK API keys, the most widely accepted integration pattern, compliant with provider terms by design.

How much does Confab cost?

Free: 1 user, 1 project, 1 channel, 7-day history. Personal: $9.99/mo for solo power users with 3 projects, /compare (2 models), and 30-day history. Pro: $49/mo for teams of 5, with 5 projects, /compare (all models), 90-day history, and full team permissions. Enterprise: $199/mo for teams of 20, with unlimited projects, unlimited history, budget controls, and priority support. Annual billing saves 20%. All plans use BYOK, so you pay your AI provider directly at standard rates with no markup from Confab.

Is there a free trial?

The Free plan is always free for 1 user with 1 project. Paid plans offer a 30-day free trial. The first 100 paid signups also get 50% off for 3 months as an early adopter discount.

What does BYOK mean?

BYOK stands for Bring Your Own Key. You provide your own API keys from AI providers like Anthropic (Claude), OpenAI (ChatGPT/Codex), or Google (Gemini). Confab uses your keys to call the AI on your behalf. You pay your provider directly at their standard API rates. Confab does not bundle AI credits, does not mark up API costs, and does not manage your AI subscription. Your keys are encrypted at rest and never stored in plaintext.

Can Confab do a SOC 2 vendor inventory?

Yes. Ask Confab to list all third-party subservice vendors your codebase depends on. The AI scans your dependencies, identifies third-party services (like Stripe, AWS, SendGrid), and commits a structured vendor inventory document to your repo. This is useful for SOC 2 Type II audits where you need to document subservice organizations.

Can Confab audit dependency licenses?

Yes. Ask Confab to analyze all dependency licenses and flag copyleft risks. The AI scans direct and transitive dependencies, identifies licenses like LGPL, AGPL, and GPL, and commits an audit document with specific packages and remediation options. This is useful for legal compliance and open-source license risk management.

Can Confab generate test plans or find coverage gaps?

Yes. QA teams can ask Confab about test coverage for any module and the AI will identify specific gaps. You can then ask it to write tests for those gaps, and it will commit them to a branch with a PR. You can also ask for a manual test plan for a new feature, and the AI will read the implementation and generate a structured plan covering edge cases.

What is the blast radius analysis feature?

Ask Confab 'what is the blast radius of PR #218' and the AI reads the diff, traces downstream dependencies, identifies affected user flows, and flags gaps in test coverage. This helps QA teams prioritize testing before manually reviewing a pull request.

Can I use Confab from my phone?

Yes. Confab lives in Slack, Discord, and Telegram, all of which have mobile apps. You open your existing chat app on your phone, message the AI, and get a codebase-aware response. There is no separate mobile app to install. Start a conversation on your phone, continue it at your desk. The chat platform handles the device switch.

How long does setup take?

Under two minutes. Sign up on the web or click Add to Slack. Connect your API keys for the AI providers you use. Connect a git repo (or describe a new idea and Confab creates one for you). Start talking to the AI in your chat channel. No infrastructure to set up, no CLI to install, no IT department required.

Does Confab store my code?

For Cloud BYOK mode, Confab clones your repo into isolated, ephemeral containers to provide codebase context. Repos are isolated per tenant and encrypted. No customer's code ever touches another customer's environment. Repo data is not used for training and is not retained beyond the configured session window.