Get a first result · GUIDE 04

Ask your first useful question

Your first question should be small enough to check. Ask the connected assistant about one note you expect it to be allowed to use, such as: “What does our client brief say about the next meeting?”

This assumes the assistant is already connected to Mensara. If it is not, ask the person who manages your setup to help.

Look for the relevant source with the assistant, then read it yourself before you rely on the answer. A source can be current or approved and still be incomplete or wrong.

What to expect

Your connected assistant may be able to see one part of your information but not another. If access is refused, ask the owner for access to the specific notes needed for this job. If the note is missing, check the locations available to your connection.

When this first question works, try a broader job only after you can recognize the source behind the answer.

Current memory and evidence · Who can read and suggest changes?

NEXT GUIDE · 05 Carry useful context into the next question Mensara can help a person or connected assistant begin with the decisions and background that matter to a job, instead of rebuilding the same briefing each time. →

Create account

MENSARA.AI · ASK YOUR FIRST USEFUL QUESTION · AGENT VIEW · TEXT/MARKDOWN

# First-read procedure

Prerequisite: owner-provided authenticated MCP setup and a principal with a granted zone. No supported client/version onboarding is verified here. Do not create a connection, provision a Mind, or use an owner identity to bypass denial.

`orient` and `list_zones` have no required arguments. `read` requires `path` and accepts `memory_view` of `current` or `evidence`; `list_files` accepts `dir` and `pattern`. Authorization is established by the MCP bearer layer, not made valid by an optional schema token field.

**Template only — placeholders come from this principal's discovery results.**

```
orient({})
list_files({"dir":"<granted directory from discovery>"})
read({"path":"<discovered map path>","memory_view":"current"})
read({"path":"<authorized note linked by the map>","memory_view":"current"})
```

After every call, parse the JSON carried in the MCP text item. For a full successful read, expect `ok=true`, `content`, `zone`, and `redactions`; ranged modes may report truncation. Cite the actual accessible path and passage. `{"ok":false,"error":"access denied"}` is a stopping condition. Do not retry it under another identity or expose the denied zone. Treat `not found` and `repository busy, try again` according to their stated condition.

The synthetic in-process Gateway examples are not MCP transport outputs. There is no established retry/idempotency rule beyond inspecting the result and retrying only an explicit busy condition.

These reads do not change the requested knowledge content. They can append signed audit receipts, so they are not free of persisted side effects. A receipt records an operation; it does not independently prove that an answer is correct.

## Continue

[Current memory and evidence](/how-it-works/current-and-evidence) · [Read and proposal permissions](/how-it-works/read-and-propose)

[Generated tool input reference](/how-it-works/tool-inputs.md)

Ask your first useful question

What to expect

Related