Memory · Episodic

Pick up where the conversation left off.

Episodic memory is your agent’s history of what happened. Record a conversation, a decision, or an outcome, then recall it with its time and source. The next interaction can build on the last one.

Episodic memory

The decision. And everything around it.

Pilot history/ September 2026
Example workspace
Original conversationPilot handoff

Monday · 10:08

A
Alex10:08

The review is complete and the pilot is ready to share.

M
Maya10:10

Great. Send it to the pilot group with the updated notes.

Outcome · source attached

The pilot moved forward after the accessibility review was completed.

Carry the history into the next conversation.

Recall the outcome and the events behind it. Bring that history into your agent’s next response.

Choose a chapter to explore

Fictional conversations illustrate how an application can store events and present their sources.

What's captured

Remember the decision and the story around it.

A customer changed their requirements on Tuesday. The team chose a new approach on Friday. Save both moments, with their sources, so the next conversation has a history to draw from.

01

Put events in order

Attach a timestamp to each event so your application can reconstruct the sequence of the work.

02

Distinguish action from outcome

Label observations, decisions, errors, and other events. Search for the kind of experience you need.

03

Know where it came from

Mark whether something was observed, reported, or inferred. A recollection keeps its source context.

04

Keep the tone

Record whether an experience was positive, negative, neutral, or mixed alongside what happened.

05

Give important moments weight

Importance scores help separate significant experiences from routine activity during retrieval and consolidation.

06

Manage what stays

Connect stored episodes to auditable forgetting, with a record of the reason for removal.

Developer example

Save a decision. Find it when it matters.

Record an investigation decision with its source and timestamp, then search for it by topic. Your application chooses which moments become part of the agent’s history.

Record a decision and search for it latertypescript
// Run server-side with credentials scoped to this Mind.
const endpoint = process.env.AKASHA_URL;
const capability = process.env.AKASHA_CAPABILITY;
if (!endpoint || !capability) throw new Error("Configure your Mind connection");

async function request(method: string, path: string, body?: unknown) {
  const headers: Record<string, string> = {
    "Content-Type": "application/json",
    "x-akasha-capability": capability!,
  };
  if (process.env.AKASHA_TOKEN) {
    headers.Authorization = "Bearer " + process.env.AKASHA_TOKEN;
  }
  const response = await fetch(new URL(path, endpoint), {
    method, headers,
    body: body === undefined ? undefined : JSON.stringify(body),
  });
  if (!response.ok) throw new Error("Mind request failed: " + response.status);
  return response.json();
}

await request("POST", "/v1/memory/episodic", {
  event: {
    id: crypto.randomUUID(), event_type: "Decision",
    content: "Investigate the Q3 revenue change by sales channel.",
    timestamp: new Date().toISOString(), source: "Quarterly review",
    context: { project: "Q3 review" }, participants: [],
    importance: 0.7, tags: ["revenue", "Q3"], related_events: [],
  },
});
const events = await request("POST", "/v1/memory/episodic/search", {
  text: "Q3 revenue", limit: 20,
});
Under the hood

An event with useful context.

Episodic memory
Record
event type · content · timestamp · source
Context
structured metadata attached to an event
Recall
search by topic or retrieve a time range
Event signals
importance · valence · source attribution
HTTP
/v1/memory/episodic · /v1/memory/episodic/search

Make the next conversation more informed.

Give your agent a history it can return to, from the first question to the final decision.