Forkable

Forkable is a customizable sales workspace for companies with customer-specific workflows.

It starts with the everyday sales surface: leads, pipeline stages, clients, projects, follow-ups, documents, auth, persistence, and row-level security. The important part is what happens next: customer-specific requests become safe, reviewable product changes that can be previewed, tested, approved, and enabled for the right account.

Product Positioning

Forkable is for companies that want their sales system to match how they sell instead of forcing every customer workflow through a generic process.

• Customize fields, stages, qualification rules, approval gates, and dashboards.
• Build customer-specific behavior on a branch before it reaches production.
• Preview changes against isolated backend data.
• Run smoke tests, then merge, deploy, and notify the requester automatically.
• Ship custom behavior behind account-level feature flags.

Demo Flow

The included demo shows a company-scoped rollout workflow:

1. Shopify requests Legal Review for deals over $50k.
2. Forkable records the request and creates a safe implementation run.
3. The change is built on a Git branch and InsForge backend branch.
4. A preview deployment and smoke-test checklist are attached to the review.
5. The developer approves the change and enables it only for the right company.
6. Stripe and other accounts keep the standard product behavior.

Stack

• Next.js App Router
• React
• shadcn/ui-style primitives
• InsForge auth, database, storage, and RLS
• Account-level feature flags
• Additive SQL migrations

Local Setup

Install dependencies:

pnpm install

Copy environment variables:

cp .env.example .env.local

Set the InsForge values in .env.local:

NEXT_PUBLIC_INSFORGE_URL=
NEXT_PUBLIC_INSFORGE_ANON_KEY=
NEXT_PUBLIC_APP_URL=http://localhost:3000

Apply migrations to the linked InsForge project:

npx @insforge/cli db import migrations/20260509181347_crm-base.sql
npx @insforge/cli db import migrations/20260509182053_forkable-approval-gate.sql
npx @insforge/cli db import migrations/20260509183048_crm-storage-policies.sql
npx @insforge/cli db import migrations/20260509190000_feature-planning-chat.sql
npx @insforge/cli db import migrations/20260509193000_agent-runner.sql
npx @insforge/cli db import migrations/20260509200000_richer-two-company-demo.sql
npx @insforge/cli db import migrations/20260509201000_customer-scoped-feature-flags.sql
npx @insforge/cli db import migrations/20260509212000_realistic-company-team-demo.sql
npx @insforge/cli db import migrations/20260509213000_scheduled-agent-tasks.sql
npx @insforge/cli db import migrations/20260509214000_remove-project-codes.sql
npx @insforge/cli db import migrations/20260509220000_company-scoped-flags.sql

Run the app:

pnpm dev

Open http://localhost:3000, sign up, then use Seed demo data from the Customize page to load the realistic pipeline and Forkable review flow.

Useful Routes

• /dashboard - pipeline overview
• /leads - lead list
• /leads/pipeline - drag-and-drop pipeline
• /feature-requests - customer customization requests
• /developer-guide - Customize page

Verification

pnpm typecheck
pnpm build

Agent Runner

The coding-agent worker lives in worker/agent-runner and is intended to run as an InsForge Compute service. It has two jobs:

1. Power the interactive feature-planning chat through Codex.
2. Poll queued agent_runs, clone the target repo, run Codex in the background, run checks, and write review artifacts back to InsForge.

Copy worker/agent-runner/.env.example to a private env file, set the server secrets, then deploy. Keep FORKABLE_AGENT_RUNNER_ENABLED=false until the Codex auth, target repo, and repo credentials are configured.

npx @insforge/cli compute deploy worker/agent-runner \
  --name forkable-agent-runner \
  --port 8080 \
  --cpu shared-2x \
  --memory 4096 \
  --region iad \
  --env-file .env.runner.local

To use a ChatGPT/Codex subscription instead of API-key billing, seed the runner with Codex's file-backed login cache on a trusted machine:

codex login
base64 -i ~/.codex/auth.json | tr -d '\n'

Put that value in your private runner env as CODEX_AUTH_JSON_B64. Treat it as a password. CODEX_API_KEY is simpler for automation, but the subscription path is useful for this demo on trusted private infrastructure.

If Compute rejects the value because it is larger than the per-env-var limit, split it into chunks and set CODEX_AUTH_JSON_B64_PART_001, CODEX_AUTH_JSON_B64_PART_002, and so on. The runner reconstructs them in numeric order at startup.

The runner also needs InsForge CLI auth for backend branches and managed deployments. Seed it from a trusted local CLI login:

base64 -i ~/.insforge/credentials.json | tr -d '\n'
base64 -i ~/.insforge/config.json | tr -d '\n'
base64 -i .insforge/project.json | tr -d '\n'

Set those values as INSFORGE_CLI_CREDENTIALS_JSON_B64 and INSFORGE_CLI_CONFIG_JSON_B64 and INSFORGE_PROJECT_JSON_B64, and set INSFORGE_PROJECT_ID to the target project id.

To route the request-planning chat through the runner, set these on the Next app:

FEATURE_PLANNING_PROVIDER=codex
FORKABLE_AGENT_RUNNER_URL=https://your-runner.fly.dev
FORKABLE_RUNNER_WEBHOOK_SECRET=your-shared-secret

Forkable uses Nia for repo context and can use Hyperspell for customer context. Index the product repo in Nia and provide the runner with:

NIA_API_KEY=your-nia-api-key

If Slack context is installed in Nia locally, the same NIA_API_KEY must belong to that Nia account/org. To mirror the local CLI config exactly, seed:

base64 -i ~/.config/nia/config.json | tr -d '\n'

as NIA_CONFIG_JSON_B64. After deploy, /health reports niaSlackReady: true and the number of Slack installations when Compute can see the Slack context.

If customer context lives in Hyperspell, also provide:

HYPERSPELL_API_KEY=your-hyperspell-api-key
HYPERSPELL_USER_ID=your-hyperspell-user-id

To activate an existing deployed runner, update its env with the target repo and keys, then set:

npx @insforge/cli compute update <service-id> \
  --env-set FORKABLE_AGENT_RUNNER_ENABLED=true