Pinlo

Use cases

What an agent
should keep.

Four workflows where a memory that survives the session changes the result: which records matter, how they come back and where the line is.

01 / Trading agents

Memory for trading agents.

Example workflow

A trading agent that restarts every morning forgets why it liked a setup yesterday. Pinlo keeps the thesis, the risk limits you set and the reasons behind each past call, so the next review starts from them.

See the workflow

Desk agent · 3 records recalled

  • ThesisOnly trade the breakout if volume holds above the 20-day average for two candles.
  • LimitsMax 2% of the book per idea; anything larger waits for a human approval.
  • Last callPassed on a similar alert on Tuesday: volume faded within the hour.

02 / Forecasting agents

Memory for forecasting agents.

Example workflow

Forecasts move as evidence arrives. Pinlo keeps each source, each estimate and each revision for a question in one mind, so an agent can explain how it got from its first number to its current one.

See the workflow

Forecast agent · 3 records recalled

  • QuestionResolves YES if the candidate wins the state; resolution source is the official count.
  • Prior estimate41% on Monday, based on two polls and early turnout data.
  • Open doubtPoll B over-sampled urban voters last cycle; weight it lower.

03 / Asset research

Memory for asset research.

Example workflow

Tokenized stocks, funds and bonds come with documents that change: terms, notices, corporate actions. Pinlo lets a research agent keep each version and each update it reviewed, and prove which ones it relied on.

See the workflow

Research agent · 3 records recalled

  • Terms v3Token tracks one share; dividends adjust a balance multiplier, not the raw balance.
  • ContractAddress and decimals read on-chain on 9 Oct; matches the issuer's published list.
  • Prior noteLast split was applied through the multiplier within one trading day.

04 / Field robots

Memory for field robots.

Proposed integration

Robots and devices learn things about the places they work: a door that sticks, a shelf that moved, a charger that fails in the rain. This is a proposed way to keep those notes in a mind that the next machine on shift can read.

See the workflow

Site robot · 3 records recalled

  • Site noteAisle 7 floor is wet after 22:00; slow to half speed.
  • HardwareDock 2 charger dropped power twice last week; use dock 4.
  • HandoverDay robot left pallet 7-14 half scanned; finish it first.

Save, recall, review

Three habits.
One record to return to.

  1. 01

    Save

    Write down observations, decisions and where they came from while they are fresh.

  2. 02

    Recall

    Pull the records a new request needs before the agent answers.

  3. 03

    Review

    Trace an answer back to the records behind it, and check them against an anchor.

Start with the API.
Grow into grants.

Read the docsOpen the app