OSNOVA — management reporting for people and AI agents
An open-source integration and management-reporting platform. Executives get answers in the moment, AI agents read the company’s financial memory over MCP — and with Sloy they also get full project context.
24/7
Management reporting in real time
Open source
A platform with no vendor lock-in and no black boxes
59 sec
From an invoice photo to a posted accounting entry
An integration and management-reporting system
02 / 09
What OSNOVA is
One financial model of the company — P&L, cash flow, chart of accounts, projects and rules — is assembled from accounting systems and is equally readable by people and AI agents.
Integration
Data from ERP, banks and accounting systems flows into one financial model with no manual entry.
For people
Reports and dashboards in the moment: a chart builder with no IT skills or ERP customisation needed.
For AI agents
The same financial memory is exposed to agents over MCP — a token is issued in one click.
Finance that speaks business
03 / 09
For executives
360° transparency
P&L, cash flow and balance sheet in the moment — not a file at month end.
Zero-IT
Dashboards and charts assembled in a couple of clicks, no training or customisation.
Smart insights
The agent finds cash gaps, expense anomalies and growth areas on its own.
Live OSNOVA interface: a chart over financial operations is built right in the product.
Financial memory that agents can read
04 / 09
Already in the product: a token is issued in one click — and the agent reads the financial memory on the user’s behalf.
AI-native financial analysis
An AI agent needs different slices of the data — and it gets them over MCP, not as "text from the internet".
"Why did project revenue deviate from plan?"
The agent compares actuals to plan and shows the cause.
"Which payments broke financial discipline?"
Overdue items, payments without grounds, amount outliers.
"What should we check before month-end close?"
A list of exceptions and missing documents.
Not just numbers — project context
05 / 09
OSNOVA + Sloy
Sloy collects Telegram, email, CRM, Google Drive and meetings into machine-readable memory. The agent answers not only "what happened" but "why": what was agreed with the client, what was discussed in chat.
30+ projects
Memory of all KT.Team client projects: sources, digests, decisions.
2–3k tokens
Linked wiki memory instead of RAG: the agent reads a project map, not the whole archive.
Any agent
Claude, Codex, Cursor — the context is available over MCP to any agentic system.
Sloy: organisation projects with connected sources (names hidden).
The agent finds a deviation and records the cause itself
06 / 09
Live case — KT.Team
AI agent
On project A two May invoices are unpaid, actuals are below plan. Asking the manager for the cause.
AI agent → manager
Hi! Project A invoices are unpaid in May. Agreed postponement or overdue?
Project manager
Agreed with the client to postpone to June 10 — we were waiting for the addendum to be signed.
Recorded in financial memory
Project A deviation: an agreed payment postponement, not overdue. The executive sees the cause in the report.
A real dialogue over KT.Team’s financial memory: an instant per-project breakdown (client names hidden).
A platform, not a box
07 / 09
Open source
OSNOVA is built on the open agent-engineering principles shaped by the industry’s best engineers — Anthropic, Google, Andrej Karpathy: transparent memory, skills, MCP, context engineering. The goal is not to sell a box that behaves unpredictably, but a platform where agents are grown with standard tools.
Standard tools
A regular user evolves agents with standard tooling — no proprietary magic.
Agents end-to-end
We build AI agents on request, turnkey — from data to production.
IT team enablement
Your team sees how the agent works and keeps evolving and maintaining it on its own.
Partnership scenarios
08 / 09
Ready to partner
Internal use
Reporting automation for your departments: a management-reporting pilot in a week, the first Sloy memory loop in a day.
External market
White-label under the bank’s brand, revenue share or an exclusive partnership for your clients.
Constructor
Adapted to your scenario: MCP access for your AI agents, on-premise or cloud.
