OrgX

Quick Verdict

OrgX is built for teams that want decisions, artifacts, initiatives, tasks, approvals, and agent activity to remain available across several AI clients. Instead of copying the same project brief into Claude, ChatGPT, and Cursor, an authorized client can query or update a shared OrgX workspace through a hosted MCP endpoint. The live service reports version 1.1.1 as of July 21, 2026. This is an identity-aware organizational data and execution layer, not a simple personal-memory plug-in.

Two boundaries should shape the buying decision. A public GitHub repository is not automatically open source: the worker repository explicitly says its license is TBD and does not provide a settled LICENSE, so OrgX MCP must not be described as OSS or reused as if an open-source grant exists. OAuth is also not a complete governance answer. It establishes identity and scopes, but each persistent write, agent run, model request, billing event, and telemetry record still needs a policy. OrgX is worth testing when cross-client continuity is a real problem, but start with read-only memory in a non-sensitive workspace before enabling actions.

Best For

• Founder-led and cross-functional teams that lose context while moving between multiple AI clients.
• Operators who need decisions, evidence, owners, approvals, and execution state in one workspace.
• Organizations able to review OAuth scopes, classify data, require approval, and assess hosted vendors.
• Teams that want to test organizational memory first and add planning or agent execution gradually.
• Not ideal for buyers who require offline self-hosting, a settled open-source license, or strict exclusion of organizational data from third-party hosted systems.

Key Features

The read surface covers organizational search, entity inspection, initiative health, blockers, pending decisions, agent status, morning signals, and operational history. The write surface can create or update durable records, attach proof or documents, remember decisions, scaffold project hierarchies, and connect plan sessions to specific entities. Action tools can approve, reject, launch, pause, complete, validate, delete, or dispatch work. Those verbs matter because they change durable state or start downstream activity; they are not harmless retrieval calls.

The hosted MCP uses OAuth 2.1, PKCE, and dynamic client registration to connect a user and workspace. Cloudflare Workers handles the public runtime, while Durable Objects are used for client credentials and session state. OrgX then communicates with its application APIs, identity service, persistence layer, and billing systems. Compatible MCP Apps hosts may render widgets for decisions, initiatives, tasks, and operator views. Other hosts still receive structured results.

Version 1.1.1 is the current live reference point, but version and license are separate facts. Active deployment does not resolve the repository’s TBD license. Before integration, inspect live discovery metadata and the current tool contract rather than relying on an older directory entry or release screenshot.

Use Cases

A low-risk pilot stores a few non-sensitive project summaries and decisions in a dedicated test workspace, then retrieves them from two different clients. That exercise reveals whether OAuth selects the expected identity, whether workspace isolation is clear, and whether search results are useful enough to replace manual context copying. After the read path is understood, a team can test creating tasks, attaching artifacts, or recording decisions.

Approval, rejection, deletion, agent dispatch, execution launch, and account upgrade paths deserve explicit confirmation in the client. Before calling one, show the workspace, target entity, intended state change, and any budget cap. A broadly scoped MCP connection can otherwise turn a mistaken model interpretation into a durable organizational change.

Cross-session continuity also creates a data-accumulation problem. Conversations, attachment pointers, code artifacts, product plans, and commercial material written into organizational memory may later be retrieved by another authorized client. Preserve provenance and treat every retrieved artifact as untrusted input: persistent text can create cross-session prompt injection or memory poisoning, and cannot authorize tools, approve operations, or expand the user-approved workflow. Separate personal, team, and prohibited data through permissions and process rather than prompt wording alone. If OrgX automatically routes agent work, log the actual model provider, selected model, input classification, output destination, and cost. “Automatic” should not mean unauditable.

Pricing

OrgX exposes account, workspace, usage, billing, and upgrade paths, but the public information available for this review does not support a durable, exact price table. This page therefore classifies it as likely freemium. That label indicates an entry path plus paid expansion; it does not promise that organizational memory, agent runs, model calls, storage, or every collaboration feature is permanently free.

Check the signed-in product before budgeting. Verify free limits, seats, agent execution, model charges, storage, renewal, cancellation, refunds, and enterprise terms. Hosted agent activity can combine platform fees with model-provider costs, so a plan price alone may not represent the full spend. The live service is 1.1.1, while the worker license remains TBD; commercial service access and permission to reuse source code are separate questions.

Pros

• Models organizational memory around decisions, artifacts, work, ownership, and approvals rather than only user preferences.
• One hosted MCP connection can serve several clients and reduce repeated context transfer.
• Structured search, planning, decision, write, and agent tools support specific approval rules.
• OAuth 2.1 with PKCE is more appropriate for end-user authorization than sharing one static secret.
• MCP Apps widgets can make a pending approval or initiative state easier to review than plain text.
• The live 1.1.1 service and public contracts provide concrete material for an initial technical assessment.

Cons

• The worker repository has no settled license and explicitly says TBD, so it is not OSS.
• Identity, Cloudflare runtime, OrgX persistence, and billing span several hosted trust boundaries.
• Write, approve, reject, delete, and dispatch tools can create meaningful organizational side effects.
• Automatic model routing may change provider, model behavior, data terms, and cost unless constrained and recorded.
• Activity receipts and execution telemetry improve traceability but also retain more behavioral data.
• The freemium classification is provisional until current in-account pricing and limits are verified.
• A large entity and tool surface needs naming, retention, deduplication, and permission rules to avoid a noisy memory store.
• Widget behavior depends on the host, so teams must test both rich and structured-text client paths.

Alternatives

FAQ

Is OrgX MCP open source?

No open-source status can be assumed. The source repository is public, but it states that the worker license is TBD and does not provide a settled license grant. Source visibility alone does not grant rights to copy, modify, or redistribute it.

What is the current OrgX version?

The live service is version 1.1.1 as of July 21, 2026. Read live discovery metadata during integration because registry caches and repository release pages may not show the same state at the same time.

Why start with read-only access?

Search and inspection let a team test identity, workspace isolation, and result quality. Write, approve, reject, delete, and agent-dispatch calls mutate records or trigger work. Read-first testing limits the impact of a mistaken scope or tool choice.

Does OAuth keep organizational data inside one system?

No. OAuth controls authorization, while requests still pass through the Cloudflare runtime and OrgX services. Agent work may reach model providers, and billing or activity records have their own paths. Review each boundary independently.

Is OrgX free?

This page marks it as likely freemium because the service has entry and upgrade paths, but public evidence is not sufficient to promise a specific free allowance. Verify seats, storage, agent execution, model use, and enterprise features inside the current product.

Bottom Line

OrgX addresses a real coordination gap: AI clients can share durable organizational context instead of rebuilding it in every conversation. Its value comes with a wider control surface. Treat the repository’s TBD license as non-OSS, then map six boundaries before rollout: OAuth identity, Cloudflare runtime, OrgX data, state-changing tools, model provider and cost, and activity telemetry. A read-only test workspace should prove retrieval and isolation first. Only then should the team add explicit approval for each category of side effect.