Skip to content
OrgLoop

Maturity Roadmap

The Goal

Section titled “The Goal”

OrgLoop’s aspirational launch story: “You can install my engineering organization. Here’s the YAML. Run this CLI.”

The manifesto ends with that promise. Everything below is the path to making it real.

Phase 1: Bootstrap Project

Section titled “Phase 1: Bootstrap Project”

Scaffolding, core runtime, CLI.

  • Monorepo setup (pnpm workspaces, Turborepo, Biome)
  • @orgloop/core — event bus (in-memory + file WAL), router, transform pipeline, logger fan-out
  • @orgloop/sdk — plugin interfaces, base classes, test harnesses for connectors, transforms, loggers
  • @orgloop/cli — init, validate, plan, start, stop, status, logs, test, add, env, doctor, routes, hook, inspect, install-service, service, version
  • YAML schema + JSON Schema validation
  • Checkpoint persistence (file-based)
  • Built-in transforms: filter (jq-based), dedup, enrich, shell script executor
  • Built-in loggers: file (JSONL), console, otel (OpenTelemetry), syslog

Exit criteria: orgloop init && orgloop validate && orgloop start works end-to-end with a mock connector.

Phase 2: Migrate Bespoke Scripts

Section titled “Phase 2: Migrate Bespoke Scripts”

Build GitHub, Linear, Claude Code, OpenClaw connectors.

  • @orgloop/connector-github — poll-based, PR activity, CI status
  • @orgloop/connector-linear — poll-based, ticket state changes
  • @orgloop/connector-claude-code — hook-based, exit notifications
  • @orgloop/connector-openclaw — webhook target, agent wake
  • @orgloop/connector-webhook — generic inbound/outbound
  • Migrate each script one at a time: build -> test -> hard cut over -> clean up
  • Launch prompt delivery (with.prompt_file) working end-to-end

Exit criteria: Every existing bespoke script has been replaced by an OrgLoop connector. The old LaunchAgent plists are deleted.

Phase 3: Operate on OrgLoop

Section titled “Phase 3: Operate on OrgLoop”

Run OrgLoop on its own org. Validate.

  • Dog-food the system: run your actual engineering org entirely on OrgLoop
  • Harden based on real-world failure modes (crash recovery, checkpoint drift, delivery retries)
  • Tune transforms, refine SOPs, iterate on the route configuration
  • Validate the success criteria: parity, latency, reliability, recovery, developer experience
  • Build observability: orgloop status tells the full story of the org’s operational health
  • Publish @orgloop/cli to npm — the first public release

Exit criteria: 30 days of stable, unattended operation. Zero dropped events. Recovery from process crashes without manual intervention.

Phase 4: Package-Native Project Model Complete

Section titled “Phase 4: Package-Native Project Model Complete”

The project model is package-native. Projects use package.json + orgloop.yaml with standard npm dependency management. A module system was briefly implemented in v0.1.8 and removed in v0.1.9 in favor of this simpler approach.

  • examples/engineering-org/ — the reference project (GitHub, Linear, Claude Code, OpenClaw)
  • examples/minimal/ — simplest possible project (webhook -> webhook)

Exit criteria: Met. orgloop init scaffolds a working project. Connectors install via npm install.

Phase 5: Launch

Section titled “Phase 5: Launch”

“Install my engineering organization right now.”

The killer demo: the manifesto ends with a live demonstration. You read the manifesto. You’re convinced. Then:

Terminal window
npm install -g @orgloop/cli
orgloop init --name my-org --connectors github,linear,openclaw,claude-code --no-interactive
cd my-org && npm install
# Set env vars: GITHUB_TOKEN, LINEAR_API_KEY, OPENCLAW_WEBHOOK_TOKEN
orgloop env                     # Verify credentials
orgloop validate                # Check config
orgloop start

Your engineering organization is running. GitHub events route to your agent. CI failures wake your supervisor. PR reviews trigger focused SOPs. Linear tickets flow through transforms. Everything is auditable, deterministic, and version-controlled.

That’s the launch. That’s what we’re racing toward.

Launch artifacts:

  • Published npm packages: @orgloop/cli, @orgloop/core, @orgloop/sdk, all first-party connectors and transforms
  • Documentation site at orgloop.ai
  • The manifesto, updated with the live demo
  • Content series: blog posts, social, community launch

Beyond Launch: Connector Maturity & the Orchestrator

Section titled “Beyond Launch: Connector Maturity & the Orchestrator”

After launch, DX deepens through two independent tracks:

Track A: Connector Maturity (organic)

Connectors progress through stages (see Scope Boundaries):

StageCapabilityUser experience
1. Functionalsource/target works”Set GITHUB_TOKEN and run start”
2. Discoverablesetup metadata, validators”GITHUB_TOKEN — create at github.com/settings/tokens”
3. Self-servicecredential acquisition”Authenticate via browser? (Y/n)”

Each first-party connector matures at its own pace. GitHub (well-established OAuth) may reach Stage 3 before OpenClaw (local service, evolving API).

Track B: Environment Orchestrator (sister project)

orgctl reads the project config and connector setup metadata to handle what OrgLoop doesn’t: service installation, credential brokering, cross-system configuration.

Terminal window
orgctl bootstrap --template engineering-org --github-repo my-org/my-repo
# Blank machine → running autonomous engineering org

See the orgctl RFP for the full project specification. orgctl depends on OrgLoop’s stable interfaces (orgloop doctor --json, --non-interactive flags, connector setup metadata) but has its own release cadence and project scope.