Use cases

Memory for
asset research.

Link issuer updates, document versions and research notes to an asset. Give research agents a record they can retrieve for the next review.

WORKFLOW EXAMPLETokenised assets
Asset research agentReview an issuer update
REQUEST

An issuer has published an update. What has changed since our last review?

Context starts here

The next response starts with what your agent already knows.

LIGHTHOUSE MEMORY03 RECORDS

Asset context

Previous document

01

Keep the content identifier of the version used in the last review.

Research notes

02

Record the assumptions and open questions from that review.

Asset context

03

Link the issuer reference and document history to the asset identifier.

Context available to recall

Example inputs and responses showing how memory fits into this workflow.

34,000Builders on Lighthouse

Saved context.
Available to connected applications.

THE PROBLEM

What gets lost
between sessions.

01

Documents lose their version

An updated link can point to different content from the document an earlier decision relied on.

02

Research starts from scratch

A new session has the asset identifier but not the issuer context or unresolved questions.

03

File access and memory get confused

Sensitive source files need explicit access handling, separate from the summaries and references used by an agent.

WITH LIGHTHOUSE

What an agent’s brain
can remember.

Give each asset a research history: document references, issuer updates and review notes. Keep the version used in a decision available for later comparison.

01 / DOCUMENT REFERENCES

Reference exact document versions

Save the content identifier of the document used in a review. Later, retrieve it and check that the bytes match the reference.

02 / ISSUER MEMORY

Retain issuer research

Save research notes, source references and open questions. Retrieve the relevant records when the next review begins.

03 / ASSET HISTORY

Link updates to earlier records

Add new records with consistent asset identifiers and tags. Compare an update with the document history and previous reviews.

04 / DATA HANDLING

Choose the right storage for sensitive data

Encrypted file storage protects source documents. For memory, Memwal uses SEAL-encrypted Walrus blobs; batched memory and index snapshots are unencrypted. Optional IPFS mirrors are public.

THE WORKFLOW

How the workflow
uses memory.

Reference the document, retrieve the earlier review, then record the update.

  1. 01

    Reference the source document

    Store the document and keep its CID with the asset identifier and review timestamp.

  2. 02

    Recall the previous review

    Retrieve relevant issuer notes, document references and unresolved questions.

  3. 03

    Preserve the new context

    Write the updated assessment while retaining the references behind earlier decisions.

WHAT LIGHTHOUSE PROVIDES

Content integrity does not establish ownership, valuation or legal validity of an asset. Those checks belong to your application.

IMPLEMENTATION

Set up the records.
Connect your application.

Choose what to store, how to retrieve it and which checks your application must perform.

Read the memory docs
01

Use a stable asset identifier in your record tags.

02

Retain the exact document version behind each review.

03

Separate file access decisions from the context passed to the model.

04

Keep ownership, valuation and compliance checks in the application.

COMMON QUESTIONS

Before
you build.

Does a CID prove ownership of an asset?

No. A CID identifies content. It does not establish legal title, valuation, authenticity of an issuer or rights attached to a token.

Can we keep earlier document versions?

Your workflow can store each version and retain its content identifier. Keeping those references lets your application retrieve the document used in an earlier review.

Are files and memory encrypted in the same way?

No. File encryption and memory encryption are separate configurations. Memwal encrypts Walrus blobs, while optional IPFS mirrors and index snapshots remain unencrypted. Batched memory is also unencrypted by default. Check the selected storage path before saving sensitive data.

What belongs in an asset research memory?

Examples include issuer notes, source references, document CIDs, review timestamps, updates and open questions. Choose records that support the next review.

BUILD WITH LIGHTHOUSE

Set up memory for asset research.