← Back to Notes & PKM
Notes & PKM by @singhcoder

agents-structured-memory

Chat-native structured memory for agents

0
Source Code

Qordinate - Structured Memory for OpenClaw Agents

Qordinate is a chat-based assistant that becomes durable, structured memory for your users and agents.

Instead of your OpenClaw agent trying to keep all state in its own scratchpad or a custom DB, it can:

  • send messages to Qordinate on WhatsApp, Telegram, or Slack
  • use a simple, structured text protocol to:
    • store facts (preferences, flags, config)
    • maintain lists (tasks, contacts, leads, etc.)
    • create reminders that Qordinate will handle over time

The agent talks to Qordinate exactly like a human would - via chat.

User Setup (what the agent should tell the user once)

Before the agent can rely on Qordinate, the user needs an account:

  1. Ask the user to open https://qordinate.ai in a browser
  2. Sign up by logging in with their phone number + OTP
  3. Start a chat with Qordinate on one channel:

Once that's done, your agent can start sending messages to that same chat window.

How the agent should talk to Qordinate

Qordinate is built around lists. Think of them as named buckets like tasks, contacts, leads, notes, etc. Your agent's job is to tell Qordinate, in plain language, what to add to which list, and what to fetch later.

At a high level, the agent should do three things:

  1. Create list items when something important appears
  2. Update list items when the status changes
  3. Query lists when it needs context or next actions

All of this can be done with natural language messages sent to Qordinate over WhatsApp / Telegram / Slack.

1. Creating items in lists

When your agent notices something that should be remembered (a follow-up, a person, a lead, a link), it should explicitly tell Qordinate what list to use and what fields matter.

Example phrasings your agent can send:

  • "Qordinate, add a new task to the tasks list: title: Follow up with ACME about the reorder pilot status: todo due: next Tuesday 3pm notes: important client meeting"

  • "Qordinate, create a contact in the contacts list: name: Harpinder Singh company: qordinate.ai role: Co-founder notes: building Qordinate"

  • "Qordinate, save this link in the resources list: title: OpenClaw + Qordinate integration idea url: https://example.com/post notes: inspiration for agent-to-agent coordination."

Key points:

  • Always mention the list name (tasks, contacts, etc.).
  • Give Qordinate a title plus a few structured fields (status, due date, notes) in plain language.
  • The agent can include where it came from (email, chat, file) so Qordinate can link back later.

How the agent should think about lists

  • Use different lists for different domains: tasks, contacts, leads, projects, resources, etc.
  • Treat Qordinate as the source of truth for "what needs doing" and "who is who" across time.
  • Whenever the agent thinks "the user will care about this later", it should create or update something in a list, not keep it in ephemeral memory.

In short: your OpenClaw agent talks to Qordinate like a slightly bossy human assistant

  • "add this to the tasks list"
  • "update that contact"
  • "show me my open tasks"

and Qordinate handles turning that into structured, durable memory.