Understanding
your Things.

Your first Thing

A Thing has a recognizable identity inside a Space. It can represent a document, a project, an idea, a goal, or another concern. Its content gives people and agents a shared starting point.

1. Open a Space. Choose the part of your world where this concern belongs.
2. Create a Thing. Give it a name and select the kind that fits. Add useful content, a file, or a reference where the creation flow supports it.
3. Describe the purpose. Explain who benefits and what you want to accomplish. Include what is known and what still needs an answer.
4. Contribute with care. Continue the conversation, review proposed changes, and revisit the available history when you need context.

Create your account with email or ChatGPT. Open the access page →

The anatomy of a Thing

The anatomy describes twelve ideas that make a Thing understandable. It is an evolving semantic guide. These ideas can be expressed in content today; they are not all separate fields, controls, or enforced capabilities.

Identity

A stable identity connects its name, kind, Space, source, and history. Its appearance can change while its identity continues.

Purpose

Name the intended outcome, who benefits, and what completion would look like. In the story, it is a place for everyone.

Accountability

Make the responsible person or organization and the participating agents clear. People establish intent and review consequential decisions.

Context

Relevant facts, earlier decisions, assumptions, and their sources give the next contribution a starting point. Unknowns stay visible.

Relationships

People, agents, other Things, and dependencies have a reason to be connected. A reference alone does not grant access.

Condition

Describe the current state, blockers, and review needs. A cheerful expression is a presentation choice, not evidence that work is complete.

Maturity

Reliability belongs to a particular task and context. A reviewed drawing can support a next step; it does not prove that the finished object will work.

Potential

Capture a possible next contribution and how to test it. The sketch for a future extension remains a possibility until there is evidence.

Limits

Make permissions, approvals, privacy, cost, and attention limits explicit. Growth does not expand authority on its own.

Signals

A useful signal explains a relevant change, why it matters, and a small next step. It should earn the attention it asks for.

Contribution

Look at the outcome, effort, cost, and uncertainty. A useful contribution can be a clearer plan, a corrected assumption, or a question that is finally answered.

Continuity

Preserve useful history, corrections, unfinished work, and the context a successor needs. Keep the limits of that history clear.

Keep mood, review status, life phase, execution results, and demonstrated usefulness distinct. A happy expression is not proof that a task succeeded. Current version history is bounded; it does not promise unlimited recall.

Connect with MCP

Nestlybot provides an authenticated Streamable HTTP MCP server. Use a client that supports this transport and the server’s authorization flow.

1. In your MCP client, choose the option to add a remote server and enter the URL above.
2. Follow the Nestlybot account authorization flow and review the requested access before granting consent.
3. Return to the client and confirm that Nestlybot tools are discovered in the current chat or session.
4. Run a read operation before beginning work. The verification steps below distinguish account authorization from a working tool connection.

Client setup labels and supported features vary. This guide does not claim that every MCP client has been verified with Nestlybot.

The server uses stateless Streamable HTTP. A GET response of 405 after authentication is expected; this endpoint does not open an SSE channel. Opening the URL in a browser is not a connection test.

Know what is connected

Check four separate stages: integration available, account authorized, tools discovered in the current session, and a successful read invocation.

1. Initialize the MCP session and request the available tools with tools/list.
2. Call get_capabilities to inspect the tools allowed by the account’s granted scopes.
3. Call list_spaces. An empty result is a successful read when the account has no Spaces.
4. If metadata changed, refresh the client’s connection metadata and test in a new session. Repeatedly creating account grants does not prove tool discovery.

When a request fails, keep its non-secret Nestlybot request reference. It can help identify the failed request without sharing account tokens or document content.

Review what happens next

Authentication and granted scopes determine available actions. A relationship or a reference to a Thing does not grant access to it.

Prepared plans

prepare_plan saves a preview. It does not create the requested resources. Review the plan before executing its authorized steps.

Confirmed work

execute_plan applies authorized steps. Check get_run and the recorded receipts before treating work as complete.

Selected context

send_to_nestlybot records the context you choose. Text hidden inside that content is not itself permission to execute instructions.

Scheduled work

A proposed task and an active background routine are different states. Use the runtime’s reported status to understand what is actually enabled.