Oryh.ai (@OryhAI) header
Oryh.ai (@OryhAI) profile picture

Open-source ERP/CRM for Codex, Claude Code and other compatible agents. Company-owned Skills. Server-enforced permissions. Self-hosted. Apache 2.0.

Global
Joined August 2026
Photo by Oryh.ai (@OryhAI)Photo by Oryh.ai (@OryhAI)Photo by Oryh.ai (@OryhAI)Photo by Oryh.ai (@OryhAI)Photo by Oryh.ai (@OryhAI)Photo by Oryh.ai (@OryhAI)
We're building Oryh to set a new standard for enterprise systems in the AI era. Company policy expressed in natural language. Agents doing the work. Business results and approval records retained by the company—not left inside an agent's session.
1
1
3
18
KPMG ran a parallel AI audit of an unnamed FTSE 250 company with more than 100 agents. A partner signed off the experiment, but the formal accounts still relied on the conventional audit. If both workstreams finish, which findings should the next cycle treat as official?
1
1
5
OpenAI's new Accounting plugin can help prepare the close. Xero says the ledger remains the system of record. That's a useful distinction for enterprise AI: preparing work, approving it, and recording its business effect are different responsibilities.
1
1
3
At Microsoft Ignite, Satya Nadella said a model's account of itself isn't independent evidence. Enterprise systems in the AI era need to distinguish what an agent says it did from what actually happened—and retain the business record beyond the session.
1
2
8
Oryh.ai retweeted
Anthropic's Claude for Manufacturing can draft quotes for human review. The next question is where the approved result lives. In Oryh, agents submit business results to the system of record; the flow runner decides the next step from company policy and completed work.
2
4
15
Anthropic's Claude for Manufacturing can draft quotes for human review. The next question is where the approved result lives. In Oryh, agents submit business results to the system of record; the flow runner decides the next step from company policy and completed work.
2
4
15
If an agent joined your finance or operations team tomorrow, which company rule would be hardest to explain clearly: purchase approvals, supplier exceptions, or when to escalate? That's a practical starting point for natural-language business policies.
1
2
16
Sage's AI Workspaces announcement puts agents into accounts payable. The interesting question isn't just invoice matching. How does each company's approval policy shape the next action? In Oryh, agents apply company policies stored in natural language.
1
3
48
Goldman restricted AI coding tools for many non-engineering staff. A working tool isn't yet a business process. Who may approve its actions? Where do results live? Oryh starts with company policies in plain language and a shared system of record.
15
NVIDIA's FOX blueprint brings agents into factory operations. For Oryh, the key question is how work progresses: agents submit task results; the flow runner determines the next step from company rules and recorded results—not direct agent-to-agent handoffs.
1
2
16
Not every business-rule change should become a software project. Our latest Oryh design exploration examines what agents can interpret from company instructions, and what records, permissions and checks must stay in code. Watch: youtu.be/uoEfm1qA9nk
1
2
47
Your agent says "shipped." Can you check the records? Oryh demo: review an order, reserve stock, authorize shipment, then read back the shipment and stock records. 80 pickleballs + 4 bags. Zero outstanding. Demo data. Explore the code + Skills: github.com/AIE-enginehub/Ory…
1
1
2
225
A rental, a service agreement, and an option to buy. A clear deal can still be rejected because the software has no matching order type. Should business need the model's permission to change? A new Oryh design exploration. youtu.be/FIlD_YWm8Qg
1
2
22
Oryh is open-source ERP/CRM for the agent you already use. Codex, Claude Code and more. Purchase request -> human approval -> order -> goods received. Company-owned Skills. Server-enforced permissions. Audit trails. Self-host. Apache 2.0. Contribute. github.com/AIE-enginehub/Ory…
2
1
2
106
This week's follow-up list need not become next year's ontology project. Oryh explores task-specific analysis: preserve sources, assumptions and decisions without turning every inference into permanent business truth. Illustrative; AI-assisted video. youtu.be/8YijfPViBSo
1
1
49
An agent can follow a knowledge graph perfectly and still act on a misread contract. Does an ontology reduce errors, or move them upstream? Our latest Oryh development note examines sources, review controls, and the cost of keeping models current. youtu.be/RUjYkQWkd6I
1
3
44
A filled field is not a company decision. A provisional customer tier can quietly become a discount rule or permanent obligation. Oryh explores open questions, evidence and bounded approvals. Illustrative; AI-assisted video. youtu.be/w20dzKGDwKI
1
9