For agents and technical readers

Find an authorised answer and verify its sources

Discover the intended Mind, follow its maps, open the sources and report only the answer they support.

Contents

Authority and discovery

Use the tools advertised by the intended connection. Call orient with {} and verify the principal, scopes and Mind. Read the relevant first_reads map before querying. The caller needs read access; this procedure does not grant it.

Example read input:

{"path":"<authorised-map-path>"}

Replace placeholders after discovery. Do not pass a token copied from another identity.

Navigate and search

Follow mapped links with read. If the map does not resolve the question, use query with the exact subject, event and time.

{"text":"<specific question including relevant entity and time>","memory_view":"current"}

For a known folder, list_files takes dir, not path:

{"dir":"<authorised-folder>","memory_view":"current"}

Open candidate records before answering. A search hit is a lead; a denial, failed read or excluded record is not a source.

Verify the answer

Match the entity, event, scope and time. Cite the source for each material claim. Label inference and unresolved conflict. For a current question, do not use an excluded historical claim as a replacement. If a source does not settle the question, state the missing evidence rather than transferring a nearby value.

One bounded follow-up

If one fact is missing, the released query contract supports one current-view follow-up. Name the facet and paths already read. Do not use this to repeat an evidence-view query or bypass an exclusion.

{"text":"<same specific question>","memory_view":"current","follow_up":{"facet":"<one missing fact>","seen":["<path-already-read>"]}}

Report an unresolved gap if the bounded pass still lacks support.

Completion and interruption

Finish with the supported answer, the source links and any precise gap. After a transient read failure, retry the same read a bounded number of times. An unavailable or incomplete result does not establish an empty area. This procedure performs no writes or external actions.

Practical guides for people and agents

Start with one useful job

Join the waitlist

Plain text for AI agents. Same content as the human page.

## For agents

- What Mensara is: it gives a team its own Mind, one shared memory for the team and their AI agents, built by Xanadu Labs LLC.
- Status: signup is not open. A person joins the waitlist at https://app.mensara.ai/#waitlist. Agents cannot create an account or sign up.
- How you can help your person: explain what Mensara does, check whether it fits a job their team repeats, and share the waitlist link.
- Agent references: https://app.mensara.ai/llms.txt lists every page as Markdown and the tool input references for connected agents.

---

# Find an authorised answer and verify its sources

Operating procedures · For agents and technical readers · [How it works](https://app.mensara.ai/how-it-works)

Discover the intended Mind, follow its maps, open the sources and report only the answer they support.

[Page as Markdown](https://app.mensara.ai/how-it-works/find-answer-procedure.md)

[Human guide](https://app.mensara.ai/how-it-works/maps-and-navigation)

## Read memory with sources

Availability: Released Core 0.20.10; discover the actual connected tools.

Who can do this: A principal with read access to the relevant area

What changes: Reads authorised records.

Limits: A generated answer needs its source checked.

## Authority and discovery

Use the tools advertised by the intended connection. Call `orient` with `{}` and verify the principal, scopes and Mind. Read the relevant `first_reads` map before querying. The caller needs read access; this procedure does not grant it.

Example `read` input:

```
{"path":"<authorised-map-path>"}
```

Replace placeholders after discovery. Do not pass a token copied from another identity.

## Navigate and search

Follow mapped links with `read`. If the map does not resolve the question, use `query` with the exact subject, event and time.

```
{"text":"<specific question including relevant entity and time>","memory_view":"current"}
```

For a known folder, `list_files` takes `dir`, not `path`:

```
{"dir":"<authorised-folder>","memory_view":"current"}
```

Open candidate records before answering. A search hit is a lead; a denial, failed read or excluded record is not a source.

## Verify the answer

Match the entity, event, scope and time. Cite the source for each material claim. Label inference and unresolved conflict. For a current question, do not use an excluded historical claim as a replacement. If a source does not settle the question, state the missing evidence rather than transferring a nearby value.

## One bounded follow-up

If one fact is missing, the released query contract supports one current-view follow-up. Name the facet and paths already read. Do not use this to repeat an evidence-view query or bypass an exclusion.

```
{"text":"<same specific question>","memory_view":"current","follow_up":{"facet":"<one missing fact>","seen":["<path-already-read>"]}}
```

Report an unresolved gap if the bounded pass still lacks support.

## Completion and interruption

Finish with the supported answer, the source links and any precise gap. After a transient read failure, retry the same read a bounded number of times. An unavailable or incomplete result does not establish an empty area. This procedure performs no writes or external actions.

## Related

- [Ask to see the source behind an answer](https://app.mensara.ai/how-it-works/maps-and-navigation)
- [Recover from a missing or poor answer](https://app.mensara.ai/how-it-works/improve-an-answer)

[Join the waitlist](https://app.mensara.ai/#waitlist)