Cards in.
Pull requests out.

Confab is a desktop app that turns the cards on your board into working code, on your own machine. Point a workspace at a repo, dispatch a card, and your own Claude builds it in a branch and opens a pull request for review.

// bring your own Claude · runs on your machine

acme-api
SR
Sarah (PM)10:14 AM
The rate limiter still returns 429 with no Retry-After header. Can you take that card?
C
Confab10:14 AM
On it. Building on a branch, right here on your machine.
Working locally
C
Confab10:22 AM
Done. Added the Retry-After header in rateLimitMiddleware.ts, self-reviewed it, and opened a pull request. View PR →
Yours, end to end
macOS Windows Your own Claude Your repo, local Foothold Foundry board Jira, Linear, Shortcut (coming soon)

AI that works with your whole team

Not another coding agent. A collaboration tool that understands your codebase, for everyone on the team, not just developers.

PMs and devs, same AI, same context

Your PM asks "how does our auth system work?" in Slack. Your AI answers with full codebase context, the same AI your developers use to write code. No terminal. No risk.

Devs get code changes. PMs get specs and answers. Everyone works in the same channels, with the same project context.

# acme-product
PM
Sarah How does our auth middleware handle token refresh?
C
Confab The refreshToken() function in auth/middleware.ts checks expiry 5 min before TTL and rotates silently...
DV
Jake Can you add a retry on 401 and open a PR?
C
Confab Done — retry logic added. View PR →

Ask them all. Compare the answers.

Can't decide between Claude's approach and Codex's? Don't. Use /compare to ask multiple AIs the same question and see their responses side by side.

Architecture decisions, code review, confidence estimation. When it matters, get more than one perspective.

# acme-product
DV
Jake /compare How should we cache API responses?
C
Claude Redis with TTL per-endpoint. Invalidate on write. Cache headers for CDN layer...
Cx
Codex Stale-while-revalidate at the edge. Cloudflare Workers KV for hot paths...
G
Gemini In-memory LRU for sub-100ms. Redis for shared state. Consider cache stampede...

Every change gets a branch. Every branch gets a review.

The AI never writes to main. Code changes, specs, research docs: everything goes to a branch with a PR. Your team reviews before anything lands.

For doc-heavy projects, turn on auto-merge and skip the ceremony. Either way, there's an audit trail.

"AI-generated PRs wait 4.6x longer for review. Feature branch throughput is up 59% but actual shipping fell 7%."
- 2026 industry research
# acme-product
PM
Sarah Can you write a spec for adding SSO support?
C
ConfabDone — created a spec covering SAML + OIDC options, migration plan, and timeline estimates.

Have an idea on the drive in? Tell your AI before you forget.

Confab lives in Slack, Discord, Telegram, and the web. Tools that already work on your phone, your laptop, and your tablet. Open the app you already have, message your AI, and get a codebase-aware answer. No extra app to install.

Your AI is wherever your chat app is. Park the car, walk in, open your laptop. The conversation is already there.

📱 Telegram
DV
Jake thought of something — we should add rate limiting before the launch. can you draft an approach?
C
Confab Drafted a rate limiting approach — token bucket per API key, Redis-backed, with configurable burst. View draft →

An execution loop you can trust

Not another chat bot. A desktop app that ships the cards on your board, on your own machine, one reviewable pull request at a time.

🌐

Your board, executed

Mark a card ready and Confab picks it up, builds it in an isolated branch, and hands you a pull request. Your backlog becomes a build queue.

🤖

Runs on your machine

A per-workspace bridge runs your own Claude locally against a repo folder. Your code and your prompts never leave your computer.

🔒

Bring your own Claude

Use your existing Claude subscription or API key. No markup, no proxy, no reselling your tokens. Compliant with provider terms by design.

📁

Every change is a pull request

Confab never writes to main. Each card lands on a branch, gets self-reviewed against its acceptance criteria, and waits for your approval.

💡

One window per project

Privacy-first Workspaces: each project gets its own window, bound to one folder, with its own board and chat. Nothing bleeds between them.

🛡

Works with your board

The board renders from a backend-agnostic provider contract. Foothold Foundry today; Jira, Linear, and Shortcut are on the way.

Use Cases

See how teams use Confab

Real scenarios, real collaboration. Not one person in a terminal. A team working together.

Sarah (PM): What changed in auth this week?
Confab: 3 changes — session rotation, PKCE flow, rate limits.
   📄 view weekly summary doc →
PM+AI

Weekly summaries without a meeting

PM asks a question in Slack, gets a plain-English summary with a link to a doc the AI committed to the repo. No terminal. No Jira export. 30 seconds.

Jake (Dev): /compare How should we handle rate limiting?
Claude: Token bucket with Redis, Retry-After headers...
Codex: Sliding window at the API gateway level...
Gemini: Leaky bucket, but consider client-side...
Dev+AI

Architecture decisions with multiple perspectives

/compare asks three AIs the same question. See where they agree (trust it) and where they diverge (dig deeper). The feature no AI provider will build.

Lisa (Founder): I have an idea for a customer feedback widget
Confab: I'll create a project for that. Name: feedback-widget?
Lisa: Yes
Confab: Done. 🔗 feedback-widget repo + concept doc
Founder+AI

Idea to project in 30 seconds

Describe an idea in chat. The AI creates a repo, writes a starter doc, and invites you to collaborate. You never touch git. Create a channel, invite your team, start building.

Priya (QA): What's the test coverage for the checkout module?
Confab: 67% line coverage. Gaps: coupon validation, multi-currency, and the Stripe webhook error path.
Priya: Can you write tests for those gaps?
Confab: Done — 5 new tests. PR →
QA+AI

Find and fill test coverage gaps

QA asks about coverage, gets a specific answer, then asks the AI to fill the gaps. No need to set up the project locally or read the test framework docs.

Alex (Designer): What component library are we using, and how is the button hierarchy structured?
Confab: Custom components in /ui. Button hierarchy: primary (indigo), secondary (outline), ghost (text-only). Sizes: sm, md, lg.
   See ui/button.tsx lines 12-48
Designer+AI

Design system questions without bothering a dev

Designer asks about the component library in Slack. Gets a direct answer with file references. No need to clone the repo or ask an engineer to pause their work.

Tom (Legal): Analyze all dependency licenses and flag any LGPL or copyleft risks
Confab: Scanned 847 deps (direct + transitive). 3 flagged.
   ⚠ license-audit.md — 2 LGPL, 1 AGPL →
Compliance+AI

Dependency license audit

Legal asks for a license risk analysis. The AI scans direct and transitive dependencies, flags copyleft licenses (LGPL, AGPL, GPL), and commits an audit doc with specific packages and remediation options.

Comparison

Confab vs. the alternatives

Most AI tools are single-user. Confab is built for teams, from the architecture up.

CategoryConfabViktorHosted OpenClawOpenClaw (DIY)Kilo
Team collaborationIndividual DMs only, no shared threadsSlack threads only
Permissions / RBACComing soon. Everyone shares everything.
Non-dev accessAgent designed for technical usersTerminal setup + config required
Channels5+ channelsSlack only23+ channels23+ channelsSlack only
Multi-model /compareComing soon
AI writes to main?Never. Branch + PR.Yes, unless configuredYes, unless configuredYes, unless configured
SecurityReverse-engineered in 2 days135K instances exposed; 12% malicious skillsYou manage it
Budget controlsComing soonNo caps. $2,400 overnight loops reported.You manage it
Bring your own AICredits. Viktor manages providers.
Cost transparencyStandard API ratesYou connect your own API keys. You pay your provider directly at standard rates. Confab never marks up your AI costs, and built-in budget caps prevent surprises.Opaque credits on top of planAPI rates applyYou manage itStandard BYOK rates
5-person team$49/mo$300–400+/mo$95–395/moFree + infra$75/mo

What to know about AI agent pricing

$

Credit-based pricing adds up fast

Some AI agent tools charge opaque credits on top of their monthly plan. Real-world reports show $200-400/month in sustained use, even when the headline price says $50. Confab charges a flat team price. No credits, no surprises.

$49/mo for five users on Confab Pro
$

Per-instance pricing multiplies with your team

Some hosted alternatives charge per running instance. Five team members means five containers. Confab charges per team, not per seat-times-instance.

Flat team pricing, not per-user infrastructure

AI provider policies are shifting

Providers are changing how they price and enforce third-party tool access. Some have suspended accounts without warning. Confab uses standard BYOK API keys, the most widely accepted integration pattern in the industry.

Standard API key BYOK, compliant by design
$

Cloud agents need budget guardrails

Agents without spending caps can loop overnight on real money. Confab ships with per-user spend caps and warnings before you hit your limit.

Built-in budget controls from day one

Confab keeps it simple. Connect your API keys from Claude, ChatGPT, Codex, or Gemini. You pay your provider directly at standard rates. We charge a flat team price with budget controls built in. No markup, no credits, no surprises.

The math speaks

Solo? $9.99/mo. Team of 5? $49/mo.

The same team on typical per-seat pricing: $75–400/mo

Pay annually and save 20%. You bring the AI. We don't mark up your provider costs.

From card to pull request.

No servers. No terminal. No IT ticket.

1

Install the desktop app

Download Confab for Mac or Windows and open it. Everything runs locally; there is nothing to deploy.

2

Open a workspace

Point a workspace at a repo folder on your machine and connect your board. Your own Claude runs inside it, against your code.

3

Dispatch a card

Drop a card into Queued. Confab builds it on a branch, self-reviews it against the acceptance criteria, and opens a pull request for you.

Security isn't a feature. It's the foundation.

Confab runs on your own machine and touches your code. We take that seriously. Every claim here is backed by an architectural decision, not a marketing checkbox.

Transparent pricing. No surprises.

Start free. Bring your own AI. We don't mark up your provider costs.

Monthly
Annual SAVE 20%

Free

$0

See what Confab can do

  • 1 user
  • 1 project, 1 channel
  • Slack or Discord or Telegram + Web
  • Ask questions, edit documents
  • 7-day session history
  • /compare, code editing
Join Waitlist

Personal

$9.99/mo

Solo power user

  • 1 user
  • 3 projects, unlimited channels
  • All channels including SMS
  • All permissions (code, docs, workflows)
  • /compare (2 models)
  • Auto-repo creation
  • 30-day session history
Join Waitlist

Enterprise

$199/mo

20 users included

  • Everything in Pro
  • 20 users (volume discounts available)
  • Unlimited projects
  • Unlimited session history
  • Per-user & org-wide budget controls
  • Priority support
Join Waitlist

Stop planning cards. Start shipping them.

Confab turns the cards on your board into reviewed pull requests, built on your own machine with your own Claude. You stay in control: every change is a branch, every branch is a pull request, and nothing lands until you say so. That's Confab.

Get Started FreeSee Pricing

// bring your own Claude · runs locally · your code stays yours