> ## Documentation Index
> Fetch the complete documentation index at: https://docs.useorgx.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Your first useful result

> Choose one path, verify the right capability, and continue from a real result.

OrgX onboarding helps you keep one useful result ready for the next session.
Start with [an example](/docs/guides/synthetic-data-trial),
[connect your tool](/docs/guides/toolkit), [save a small draft](/docs/quickstart), or
[build with the API](/docs/api/quickstart). Install only what your chosen path needs.

## The web journey

| Stage | Your action | What OrgX can claim |
| - | - | - |
| Outcome | Describe one client deliverable or team outcome | A brief from you; connected evidence and missing sources are labeled |
| Plan | Edit scope, deliverables, and an acceptance check | A proposed draft, with proposed agent owners and you as reviewer |
| Access | Choose read preferences and action boundaries | Your policy preference; provider authorization remains separate |
| Connect | Choose a browser connector or one local tool | A permitted read or current runner readiness; capture remains separate |
| First proof | Open evidence for this initiative | A matching artifact or decision is available; acceptance happens on its record |

No connected sources are required to describe an outcome. When discovery finds
none, the plan remains a proposal from your brief. External deep search is an
explicit choice with provider cost and limitations; it is never enabled as an
implicit fallback. Draft creation and planning-model use do not prove a run started.

Your progress is saved. If work is waiting on a run or approval, open this
initiative in Mission Control and return later. Retry the saved step when a
save receipt is missing; do not start another journey to recover.

## Choose the tool you already use

| Client | First setup | First useful read | Boundary |
| - | - | - | - |
| ChatGPT | [Manual developer or workspace setup](/docs/guides/chatgpt-integration) | Bootstrap when discovered, then retrieve a linked decision | OAuth; no local Node requirement; catalog approval is distinct from publication |
| Claude web / Desktop | [Custom connector](/docs/guides/cursor-mcp-setup) | Bootstrap when discovered, then retrieve one record | Browser OAuth; a connector is separate from a machine runner |
| Claude Code public plugin | [Plugin README](https://github.com/useorgx/orgx-claude-code-plugin) | Search, initiative pulse, or chronicle | Seven-tool `claude-directory` status profile; no automatic hooks or write tools |
| Claude Code operating connector | [Direct MCP or Wizard](/docs/guides/cursor-mcp-setup) | Bootstrap when discovered | Write access is a separate grant; a runner needs its own readiness probe |
| Codex | [Plugin README](https://github.com/useorgx/orgx-codex-plugin) or [Wizard](/docs/guides/wizard-cli-onboarding) | Confirm workspace and retrieve one linked record | Runtime installation and Wizard-owned session capture are separate |
| Cursor / VS Code / Windsurf / Zed / Gemini | [Local client guide](/docs/guides/cursor-mcp-setup) or [Wizard](/docs/guides/wizard-cli-onboarding) | A permitted record read in the host | Wizard requires Node 22+; restart or refresh the host after config changes |
| OpenClaw | [Setup guide](/docs/guides/openclaw-plugin-setup) | The documented read-only validation | Pairing, plugin health, execution, and sync delivery are separate checks |
| OpenCode | [Plugin README](https://github.com/useorgx/orgx-opencode-plugin) | Confirm workspace using discovered tools | Alpha package; do not assume production or hook parity |
| Grok Bot | [Plugin README](https://github.com/useorgx/orgx-grokbot-plugin) | A discovered context read | Hook coverage is not guaranteed; inspect the installed integration |
| DeepSeek / Goose / other advanced hosts | [Wizard reference](/docs/guides/wizard-cli-onboarding) | A permitted read from the selected host | Host support varies; do not infer a runtime or capture capability from MCP config |

Verify only tools present in the selected profile and current grant. A narrower
profile cannot acquire missing tools by following a generic prompt. An empty
workspace is a valid result; show the empty state and its next action.

## Readiness has several meanings

| State | Required evidence | Does not imply |
| - | - | - |
| Local configuration saved | Configuration inspected in the selected host | Hosted OAuth works |
| Read verified | Host retrieves expected workspace data through an allowed tool | Writes or machine execution are authorized |
| Runner ready | Fresh server presence and readiness probe for this user and workspace | Every run is approved or every session is captured |
| Capture configured | Capture mode and host targets inspected | A receipt reached OrgX |
| Evidence received | Durable record belongs to this journey's initiative | Human acceptance |
| Accepted result | A person's acceptance record for the artifact and criteria | All later revisions are accepted |

Browser pairing events only prompt rechecking server state. They cannot prove
readiness or survive as authority across accounts, workspaces, revoked keys, or
stale attempts. A saved portfolio entry is not a started run.

## Recovery

| Symptom | Next action | Keep |
| - | - | - |
| Endpoint unreachable | Check network and hosted service health | Saved plan and client selection |
| Local config missing | Inspect `surface list` and the host config; add only the selected integration | Other client config |
| OAuth missing or expired | Reauthorize inside the selected host | Workspace and intended profile |
| Wrong workspace | Reconnect the intended workspace before any write | Returned IDs; never guess replacements |
| Tool absent | Inspect discovery and the profile; choose operating access separately if needed | The current read-only connection |
| Runner not ready | Check provider sign-in, runtime, transport, and policy in Settings → Execution | Draft and approval boundary |
| Receipt delivery pending | Inspect capture configuration and delivered records separately | Local capture records and retry context |
| Save receipt missing | Retry the saved mutation from its recorded version | Outcome, request, and idempotency identity |

Changing the discovery mode, full goal, sources, or workspace creates a new
request identity. Retrying the same request preserves it.

## Data and credentials

Before installing, review what changes locally, what data can be sent, the
capture mode, and how to remove or revoke access. Show essential boundaries
before setup; expand the reference only when needed.

| Path | Authorization | Remove or revoke |
| - | - | - |
| Hosted MCP | OAuth in the selected host; profile and verified scopes intersect with workspace access | Disconnect and revoke the remote grant; reauthorize for a changed scope |
| Wizard | Local auth store or configured environment; inspect the resolved source with `auth status` | `auth clear` clears local saved auth; remote revocation is separate |
| Local plugin / runner | Its documented installation key and provider login | Uninstall integration and revoke its OrgX key independently |
| REST | Server-side `ORGX_API_KEY` | Revoke the key in OrgX Settings |

A credential removal is not deletion of hosted records. Capture metadata can
include project names or paths. See [security and retention](/docs/resources/security).

## Specialized journeys

Team Trail invitations join a specified Trail team and its pooling scope;
they do not grant workspace-wide membership or access. Expired or used links
need a replacement invitation from the team owner.

X-Ray, self-audit, and Workload Doctor begin from their own evidence or workload
shape. Workload plan confirmation acknowledges a proposal; account creation
and authority grants remain separate. Studio onboarding configures a brand
from the supplied URL or brief; its success is a usable Studio configuration,
not an accepted organizational work result.

Assisted founding-customer onboarding follows the signed agreement. Recording,
check-ins, Day-14 brief delivery, refund terms, and next-quarter plans are
customer-specific rather than promises of the default web journey.

## Verification boundaries

Source contract reviewed September 30, 2026: Wizard 0.1.85 (Node 22+), MCP
manifest 1.1.5. Source review is not a fresh package install or a deployed
OAuth negative control. Record the tested release, host version, workspace,
profile, permitted read, denied write, runner probe, and delivered receipt
before calling an installation end-to-end verified.

`onboarding_completed` is navigation graduation. It is not accepted first
value. Work milestones must be joined to the workspace, journey, initiative,
run or artifact, and intent; track readback activation separately for status
clients. A receipt marked completed can describe a completed API operation
while its task remains `todo`.