Scattered routines
A Claude Code script here, an n8n flow there, an agent someone built one afternoon. Nobody knows how many there are, or which ones still work.
Memory for Technology Companies
Scripts, flows and agents built by each team go into one catalogue, with the cost, status and history of every run. And any AI tool queries the same company memory, over MCP or the API.
The problem
And each tool knows the company in a different way.
A Claude Code script here, an n8n flow there, an agent someone built one afternoon. Nobody knows how many there are, or which ones still work.
Each team has its own key and its own subscription. AI spend only shows up at the end of the month, spread across several providers.
Engineering uses Claude, support uses ChatGPT, product uses Gemini, and each person explains the company to the AI from scratch.
Before and after
From loose routines to a catalogue with cost and versions.
| Without the company's memory | With Kluco |
|---|---|
| Without the company's memoryAI routines running in scripts and flows nobody keeps track of | With KlucoEach routine registered as an external agent, with the cost, summary and errors of every run |
| Without the company's memoryNobody notices when a routine stops running | With KlucoThe catalogue shows whether each routine is on time, late or failing |
| Without the company's memoryA change to an agent goes straight to the people using it | With KlucoA draft kept apart from the published version, with regression tests before publishing |
| Without the company's memoryEach AI tool has its own context | With KlucoChatGPT, Claude, Gemini and Copilot query the same memory over MCP |
| Without the company's memoryAI spend shows up on each provider's bill | With KlucoAI investment shows up in Brazilian reais, per team, with the month's forecast |
What goes into memory
Documents come in from Google Drive, OneDrive, SharePoint, file upload (Markdown included), API or webhook, along with resolved support conversations and recorded meetings, and become articles that AI keeps up to date.
Use cases
The most common use cases in technology companies, with the part of the platform that does each one.
Routines that run in Claude Code, n8n or scripts report every run, with cost, summary and errors, and can have a budget.
AgentsChatGPT, Claude, Gemini and other MCP-compatible tools query the same knowledge base, each person with their own access.
BrainEach group of people uses the models, agents, skills and memory its profile allows.
LLM Management and ChatEach team's AI spend, in Brazilian reais, with the month's forecast and suggestions for cheaper models.
LLM Management and ChatThe steps of a process, the criteria of an analysis or the tone of an answer are kept as a skill, with versions. Agents and people use the same skills.
BrainThe agent serves customers on WhatsApp and the other channels with what is in the reviewed knowledge base, and hands over to a person what it does not know.
AgentsAgents · AI agents
Agents built on the platform and routines that run outside it, such as Claude Code, n8n and scripts, appear in the same catalogue. External routines keep running where they are and report every run, with cost, summary and errors.
Brain · the memory
The Brain turns runbooks, policies and documentation into articles with an owner and a review deadline. Through the MCP server and the API, ChatGPT, Claude, Gemini, Copilot and the company's agents query the same knowledge base. Support varies from one tool to another.
LLM Management and Chat
Engineering, support and product use the models their profile allows in the Chat, from OpenAI, Anthropic, Google Gemini, DeepSeek, Grok and OpenRouter, with the keys set up in one place.
Control
The company sets up each control below in the console.
Related
FAQ
No. Register each routine as an external agent: it keeps running where it is (scheduled task, script, Claude Code, n8n) and starts reporting every run, with cost, summary and errors. The catalogue shows whether it is on time, late or failing, and the month's budget can warn or pause the routine.
The ones that connect over MCP, the open industry standard, such as Claude, ChatGPT, Copilot and Gemini. Support varies from one tool to another. Each person signs in with their own access and reaches only their company's knowledge base, and internal systems connect through the API.
Through Google Drive, OneDrive, SharePoint, file upload (PDF, Word, Excel, CSV and Markdown), API or webhook. The documents stay where they are, and the Brain writes articles from them, always with a reference to the source.
Yes. Service agents answer on WhatsApp and other channels with what is in the knowledge base, and hand over to a person what they do not know. Each article defines whether every agent can read it or only internal assistants, so agents that talk to customers do not read internal articles.
Next step
In the demo, we show the agent catalogue with an external routine, the memory queried over MCP and AI costs in Brazilian reais.