AchiralAchiral

Docs · LLMs and developers

Version3.14.0

Find each /memory section, pick a scope, and see which roles can read it.

Memory pages

The Memory pages show what Achiral has learned for your organization, for you, and for your agents. They live at /memory on your workspace host, https://{your-slug}.achiral.ai. Every member can open them.

Sections

SectionPageWho can open it
Overview/memoryEvery member
Top of Mind/memory/top-of-mindEvery member
Core Memories/memory/coreEvery member
Self Reflection/memory/reflectionOwners and admins
Habits/memory/habitsEvery member
Ledger/memory/ledgerEvery member
Operations/memory/operationsOwners and admins
Agents/memory/agentsEvery member

A member's section menu leaves out Self Reflection and Operations. Opening one of them as a member sends you back to /memory. /memory/overview also goes to /memory. The menu ends with Settings for owners and admins and Account for members.

  • Overview shows dials and counts for the scope you pick.
  • Top of Mind lists the scope's memories, most activated first, 25 at a time. Load more fetches the next page.
  • Core Memories for the organization holds the Memory API write mode and the queue of core-memory candidates waiting for approval or rejection. Your own Core Memories shows your Learned Preferences, your EA's EA Observations, and the candidates waiting for your review.
  • Self Reflection shows and runs the reflection loop.
  • Habits for the organization lists the environment's habits with their scores. Your own Habits shows your automation rules.
  • Ledger for the organization lists its memories. Owners and admins can search, add, correct, pin, let fade, prune, restore, and export them. Your own Ledger has the Memories, Timeline, Sharing, and Recall settings tabs, and Export.
  • Operations summarizes the reasoning jobs that read the workspace memory. It appears once the workspace memory is ready.
  • Agents lists the agents you can manage. While an Agents page is open, the menu shows New agent (/memory/agents/new), Runtime (/memory/agents/runtime), and Docs, which opens the agent docs, under it.

Scopes

Overview and Top of Mind have a scope picker beside the section heading. The chosen scope is kept in the page address as ?scope=, so a link opens the same view.

  • Organization, labeled with your organization's name: the organization's shared memory. Memory that belongs to an agent is not included; it appears under that agent.
  • You, labeled with your name: your own memory. If you have a personal EA, a second checkbox labeled with your EA's name sits beside yours. Keep both checked for everything, or clear one to see only your part or only what your EA learned.
  • Another member: owners and admins pick a person from the list beside their own name. Another member's memory shows counts only, never contents. This scope is offered on Overview, not on Top of Mind.
  • Agents: pick one agent from the list. Manage agent opens it in Memory > Agents. The button appears when there is at least one agent you can see.

Core Memories, Habits, and Ledger show either the organization's memory or your own. Owners and admins pick between the two beside the section heading; members see only their own. Self Reflection and Operations always show the organization's.

Owners and admins open on the Organization scope; members open on their own.

Who can see which scope

ScopeMemberAdminOwner
Organization, on Overview and Top of MindYesYesYes
Organization, on Core Memories, Habits, and LedgerNoYesYes
Your own, with your EAYesYesYes
Another member (counts only)NoYesYes
AgentsAgents they createdAny agentAny agent

The organization memory API checks each scope again, so a scope that isn't offered to you can't be read by changing the address.

Overview dials

Every scope shows the dials of its own memory:

  • the Memory Pressure, Coherence, Drift, and Memory Depth gauges;
  • the mix of episodic, semantic, and foundational memory, with a Top of Mind preview beside it;
  • Stored, Foundational, and Pending candidates counts; and
  • the Declarative Memory, Memory Chunks, and Flashbulb / Core modules.

These measure the whole workspace, so only the Organization scope shows them:

  • the Compounding Readiness gauge;
  • the Habits and Messages counts and the Procedural Habits module; and
  • the Active members and Soak time tiles.

Another member's scope carries counts only, so its Top of Mind preview stays empty.

Write mode and the endpoint

The Memory API write mode is set on Core Memories (/memory/core), beside the candidate queue it feeds. Owners and admins can change it.

The Memory API base URL and memory endpoint are shown only in Settings > API (/settings/api), next to the API keys, for owners and admins. /memory/endpoint redirects there.