Skip to content
Afara
Afara CLI is in public beta


in your coding agent.

Afara turns each team's code and tickets into context, and injects it into every coding agent that builds on it.

From other teams

Backend

payments-api, as built at 0113454

  • POST /v1/payment-methods returns a method ID
  • POST /v1/intents/confirm can return requires_action
  • A declined card is retried once on the server
Product

PAY-981, Card payment with 3-D Secure

  • Show the bank challenge when it asks for one
  • Show a clear error after a decline
Enterprise pattern

Web clients call through a BFF

  • checkout-web talks to checkout-bff, never to services
injects

In your coding agent

Frontend engineer, checkout-web

Build against

  • checkout-bff: create payment method, then confirm the intent
  • requires_action: open the 3-D Secure challenge

Don't

  • Retry a declined card in the client. The API already retries once.

Done when

  • The challenge appears when the bank asks
  • A decline shows the error from PAY-981

Works with Claude Code, OpenAI Codex CLI and Gemini CLI.

Works with the agents and tools your teams already use

Each team builds on what the others already know.

Today that context lives in other people's heads, repositories and boards. Afara puts it where the work happens: in the agent.

Frontend

Backend implementation and their own tickets

Endpoints, responses and edge cases as the backend actually built them, next to the ticket being worked on.

QA

Frontend and backend context

Every screen, call and server-side branch, so the test plan covers the app that exists.

Engineering

Product and business context

Acceptance criteria and business rules, in the agent, while the service is being written.

Everyone

Your company's patterns

How services communicate and how each one is built, applied by every agent on every team.

Patterns

Set the rules once. Every agent follows them.

Your company writes down how it builds software. Afara injects those patterns into the local coding agent whenever a service or feature is built.

Enterprise patterns
How services communicate across the organisation: protocols, messaging and entry points.
Service patterns
How each service is built inside: layers, boundaries and where business rules live.
Patterns for enterprise

Enterprise patternrequired

Web clients call through a backend for frontend

  • Browsers and apps never call domain services directly
  • State changes between services are events, not calls

applies to: checkout-web, checkout-bff, payments-api

Service patternhexagonal

Business rules live in the domain layer

  • Adapters for Stripe, Postgres and Kafka sit behind ports
  • Controllers hold no business logic

applies to: payments-api

Context that comes from the code, not a wiki.

Afara sits next to git. Context is derived from commits as teams ship, so it is never written by hand or left to go stale.

$brew install --cask radicand-labs/tap/afara

Or install with curl or Scoop.

  1. Generate

    Capture what each team built

    Your AI coding tool reads new commits in a read-only copy of the repository and draws the feature: screens, steps and API calls, each tied to a file and line range.

    $ afara generate --feature "card payment"
  2. Link

    Attach the ticket and its intent

    Link the Jira, Linear or GitHub issue. Afara reads the story into the same kind of graph, so product context travels with the code.

    $ afara link --issue PAY-981
  3. Push

    Share it with the organisation

    Only the derived context leaves your machine. It becomes available to every team that builds on this feature.

    $ afara push
  4. Build

    Every agent starts with the full picture

    When another team builds on the feature, Afara injects its context, their tickets and your company's patterns into their local coding agent.

Context you can trust, because it's checked.

Afara compares the ticket with the code and flags behaviour nobody wrote down, so no agent builds on a story that is out of date.

Card payment with 3-D SecurePAY-981
  1. Review cart

    0.94

    Cart review

  2. Enter card details

    0.91

    Card form

  3. Tokenise the card

    0.88

    Create payment method

    POST /v1/payment-methods

  4. Confirm the payment

    0.90

    Confirm payment intent

    POST /v1/intents/confirm

  5. not in story
    extra

    Retry authorisation

  6. 3-D Secure challenge

    0.83

    3DS challenge

  7. Show confirmation

    0.92

    Order confirmed

  8. Email the receipt

    missing
    not built
screen stepnumbers are alignment confidence

Your source code never leaves your machine.

Teams share context, not code. The analysis runs where the code already lives, with the AI tool you already use.

How we handle your data
  • Commits, never your working tree

    Uncommitted changes are never read or sent anywhere.

  • Analysis on your machine

    Your own AI tool reads the code in a read-only copy. Afara never clones it.

  • Only derived context is shared

    Wireframes and references to files and lines. Source stays where it is.

  • Passwords stay out of the terminal

    The CLI signs in through the browser and keeps a key only you can read.

Latest writing

See Afara on one of your own features.

A 30-minute walkthrough with an engineer. Bring two repositories and a ticket, and see the context your agent would get.