A history you can query.
Connect facts and revisit how they changed. See what happened and when your agent learned about it.
Give your AI agents a place to remember, connect knowledge, and build on past work. Keep the context when the conversation ends.
Your own dedicated instance. Start free.
Made to remember.
The decisions. The details. The thread of an idea.
Six kinds of memory. One continuous Mind.
Meet Helm / The Minds Operator
Start a dedicated Mind through a conversation with Helm.
Clear choices, a visible action, and a result you can inspect.
Your workspace. One conversation away.
Talk to HelmOne engine
Akasha, the engine behind Minds, brings records, similarity search, connected facts, and SQL analytics together. Give your application several ways to use the information it keeps.
The join path connects kv, vector, graph, sql through one Minds engine to joined context. This diagram illustrates a data flow; it does not display live traffic.
Use SQL to ask structured questions across supported data sources.
Illustrated flowA Mind connects what your agents remember with how they reason, learn, and act.
A past conversation is different from a password or a reusable workflow. Minds gives each kind of information a place, so your agents can use it for the right task.
The team chose one shared workspace for the first release.
Record the events, decisions, and outcomes that matter. Your agent can return to saved experience and the sources you attach.
Give your agent a way to plan smaller steps and connect pattern matching with rules and known relationships. Minds provides reasoning tools you can configure around your task, with paths you can inspect.
Connect people, projects, and events so your agent can follow relationships. When a fact changes, keep both when it changed and when the update arrived.
A spiking neural network processes signals over time. Build applications that look for patterns in sensor readings, market ticks, or logs, where the timing matters.
The platform at a glance
Experiences, workflows, resources, secrets, working context, and patterns.
Records, vector search, a knowledge graph, and SQL analytics in one engine.
Your own instance, with the same Mind when you upgrade from Free to Pro.
Talk to Helm, connect an SDK, or work from the command line.
Useful memory has to return the right context. Explore how to inspect retrieval and understand what an evaluation measures, from the questions asked to the method used to judge the results.
Knowledge lineage records how a claim was formed and which sources contributed to it. Follow that trail to inspect the evidence before you rely on the result.
Original records
Identify the evidence
Evaluate the result
Keep the source trail
Depth 3 · Derivation · 4d ago
The revenue claim completes this illustrated trace. Follow its derivations back to the interview and the filing that contributed to it.
Bring a saved decision into the next request. These examples search existing memories from your application or terminal. The HTTP examples use a Mind with MCP enabled. Set AKASHA_URL to its endpoint, AKASHA_TOKEN to its bearer token, and AKASHA_CAPABILITY to an issued capability token authorized for the operation.
const capability = process.env.AKASHA_CAPABILITY;
if (!capability) throw new Error("Set AKASHA_CAPABILITY before connecting");
const response = await fetch(
new URL("/v1/mcp/tools/call", process.env.AKASHA_URL),
{
method: "POST",
headers: {
Authorization: "Bearer " + process.env.AKASHA_TOKEN,
"x-akasha-capability": capability,
"Content-Type": "application/json",
},
body: JSON.stringify({
name: "recall",
arguments: { query: "What release schedule did we choose?", limit: 5 },
}),
},
);
if (!response.ok) throw new Error("Memory request failed: " + response.status);
const output = await response.json();
if (output.is_error) throw new Error("Memory tool failed");
console.log(output.result.memories);Choose a compute size, how much data you keep, and how much you transfer each month. Free has capped usage. Pro starts at $29; larger sizes and usage beyond the included amounts add to the cost.
Free includes the Micro compute size. Upgrade to Pro when you need more capacity.
A dedicated Micro instance with capped storage and transfer.
Every Mind is dedicated. This estimate covers compute size, storage, and transfer; capability selections do not change the estimate. Model-provider usage and separately contracted integrations are outside this estimate. Engine services may require configuration or training.
Start with a dedicated Mind. Upgrade to Pro for more capacity, with pricing based on compute, storage, and transfer. Engine access is included; model-provider usage and separately contracted services are outside these resource prices.
Keep your knowledge in a dedicated Mind. Read how Minds processes account and service data, and how connected providers may handle your requests.
Built for the work
“Context changes
the answer.”
Source
passage
Linked
citation
Research
Keep the papers, passages, and sources behind a finding. Return to the evidence when a question comes up later.
ExploreRefund resolved.
Following up.
Customer memory
Bring past conversations and useful resolutions into the next interaction, so your agent can respond with the customer’s history in mind.
ExploreNew run. The work carries on.
Autonomous agents
Save decisions, working context, and reusable steps. Help your agents pick up a project without rebuilding the context each session.
ExploreA useful conversation should leave something behind: a decision, a source, a better way to do the work. Give your agents a Mind to carry that context into whatever comes next.