Pre-launch — building now

AI won’t transform your business. Your data will.

Every sales and success call holds what customers really think. It’s all recorded somewhere. Almost none of it gets used. Miruka makes that hidden intelligence visible, as clean, private context your agents can use, and it stays yours.

What you’re not using

You have the answers. You just can’t see them.

Conversations.

Sales calls, CS emails, meetings. Your reps hear the objection. Your CRM never does.

Disconnected systems.

Your CRM doesn’t speak to your accounting software. The pattern sits between them, unseen.

No public AI model has ever seen this data. That’s your edge.

How Miruka works

  1. Capture.

    Miruka listens to your sales and customer-success conversations and connects them with your systems.

  2. Clean.

    Noise depends on the task. What an SDR agent needs isn’t what a campaign agent needs. Miruka agents remove the noise for each job.

  3. Serve.

    Any agent, built on any LLM, gets only the signal it needs, through MCP, with your permission.

Example signal, for an SDR agent.

Better answers. Lower agent costs. Agents read the signal, not every raw row.

Your second brain

Your knowledge stays your IP.

One brain per company.
Yours alone.
Never used for training.
We don’t train our models on your data.
Never shared.
Not with other clients. Not with public models.
LLM-agnostic.
Works with the agents you already use, or will.

見える化

mieruka — “making the invisible visible.”

At Toyota, mieruka meant putting problems where everyone could see them, so they got solved. We’re applying the same discipline to what your customers tell you.

# Miruka
 
> Miruka is a private second brain for the revenue teams of SMEs ($10M–$100M in revenue). It is being built. Nothing is live yet.
 
Miruka makes hidden customer intelligence visible. It captures what customers and prospects say in sales and customer-success calls, emails and meetings, and connects it with data from systems that don't talk to each other, such as CRM and accounting. Miruka agents remove the noise for each specific task, so other agents get only the signal they need. Agents do better work and cost less to run because they don't read every raw row. The cleaned knowledge stays the company's IP.
 
## product
name:             Miruka
category:         private context layer for AI agents in revenue teams
for:              sales, customer-success and marketing teams at SMEs with $10M–$100M revenue
origin:           mieruka, "making the invisible visible", a Toyota practice
 
## status
status:           building
stage:            pre-launch
live:             false
 
## capabilities_planned
capture:          sales and customer-success conversations (calls, emails, meetings), connected with CRM and accounting data
clean:            task-specific noise removal by Miruka agents; an SDR agent and a campaign agent need different signal
serve:            only the signal each agent needs, to agents built on any LLM, through MCP, with the company's permission
 
## data_policy
tenancy:          one brain per company
training:         never; client data is not used to train models
sharing:          never; not with other clients, not with public models
ownership:        the cleaned knowledge remains the client's IP
llm:              agnostic
 
## access
mcp:              coming (not live)
api:              none yet
 
## contact
follow_the_build: POST /api/follow with JSON {"email": "<address>"}
humans:           the email field on / labelled "Follow the build"
 
## files
llms:             /llms.txt
agent_card:       /.well-known/agent-card.json
 
# /.well-known/agent-card.json
 
{
  "protocolVersion": "0.3.0",
  "name": "Miruka",
  "description": "Miruka is a private second brain for the revenue teams of SMEs ($10M–$100M in revenue). It is being built. Nothing is live yet. Miruka makes hidden customer intelligence visible. It captures what customers and prospects say in sales and customer-success calls, emails and meetings, and connects it with data from systems that don't talk to each other, such as CRM and accounting. Miruka agents remove the noise for each specific task, so other agents get only the signal they need. Agents do better work and cost less to run because they don't read every raw row. The cleaned knowledge stays the company's IP.",
  "url": "https://miruka.ai/",
  "version": "0.0.0-building",
  "provider": {
    "organization": "Miruka",
    "url": "https://miruka.ai/"
  },
  "documentationUrl": "https://miruka.ai/llms.txt",
  "capabilities": {
    "streaming": false,
    "pushNotifications": false
  },
  "defaultInputModes": [
    "application/json"
  ],
  "defaultOutputModes": [
    "application/json"
  ],
  "skills": [],
  "x-miruka": {
    "status": "building",
    "live": false,
    "capabilities_planned": {
      "capture": "sales and customer-success conversations (calls, emails, meetings), connected with CRM and accounting data",
      "clean": "task-specific noise removal by Miruka agents; an SDR agent and a campaign agent need different signal",
      "serve": "only the signal each agent needs, to agents built on any LLM, through MCP, with the company's permission"
    },
    "data_policy": {
      "tenancy": "one brain per company",
      "training": "never; client data is not used to train models",
      "sharing": "never; not with other clients, not with public models",
      "ownership": "the cleaned knowledge remains the client's IP",
      "llm": "agnostic"
    },
    "access": {
      "mcp": "coming",
      "live": false
    },
    "contact": {
      "method": "POST",
      "endpoint": "https://miruka.ai/api/follow",
      "body": {
        "email": "<address>"
      },
      "purpose": "follow the build"
    }
  }
}