Connectors
REST APIs, databases and MCP servers all register as tools. Import an OpenAPI spec and every endpoint becomes something the agent can call.
KSABuilt for enterprises in the Kingdom
AI agents that work inside your systems
Zamil connects your ERP, databases and internal APIs to one AI layer. Your team asks in plain language, and the work happens in the systems you already own.
- Runs on your cloud or ours
- Approvals before sensitive actions
- Every action logged
Your systems
ERP
CRM
HR
Finance
Databases
Internal APIs
Zamilagent coreYour people
Portal
Team chat
Your own app
Partners & Enablers
Backed by the best in the Kingdom.
How it works
What it takes to get an agent working
There is no SDK to learn, and the system you are connecting does not change.
- 01
Connect a system
Point Zamil at an OpenAPI spec, a database, or an MCP server. It reads the schema and registers each action as a tool, with credentials injected at call time rather than written into a prompt.
- 02
Set the rules
Decide which actions need a human signature. Sensitive calls stop and wait for an approver you nominate: the person who asked, the agent's owner, or anyone on the team. Everything else runs.
- 03
Put it to work
Someone asks a question in plain language. The agent breaks it into steps, runs the independent ones in parallel across systems, and answers with a record of everything it touched.
Platform
The parts you would otherwise build yourself
Connector plumbing, planning, retrieval, approvals and audit ship with the platform.
Task planning
The model turns a goal into a dependency graph at run time and runs the independent branches in parallel. Nothing is hard-coded, so the plan changes when the question does.
Your documents
Retrieval across contracts, policies and manuals with inline citations, so an answer points back to the page it came from.
Human approval
A sensitive tool call pauses until an approver responds. The verdict, who gave it and when are stored with the run.
Guardrails
Credentials encrypted at rest, isolation between organisations, and input checks that stop a request before it reaches the model.
Admin and usage
Usage, token spend and connector activity per team, so finance can see where the budget goes.
Systems
If it has an API, a schema or an endpoint, it connects
Zamil talks protocols rather than vendors. The names below are examples of what teams point it at, not a fixed list.
REST and OpenAPI
Import a spec and every endpoint becomes a tool. Works the same for a vendor platform and for the internal system your team wrote in 2011.
- SAP
- Oracle
- Salesforce
- ServiceNow
- Microsoft 365
- Custom APIs
Databases
Zamil reads the schema, annotates it, and queries inside the permissions you grant the connection.
- PostgreSQL
- MySQL
- Oracle DB
- SQL Server
MCP servers
Point Zamil at an MCP endpoint and its tools appear next to everything else, under the same approval rules.
- Model Context Protocol
Messaging channels
Results and approval cards go out to your team's chat, so people act where they already work.
- Approval cards
- Task notifications
Governance
Built to survive a security review
The questions your CISO will ask, answered in the product rather than in a policy document.
- Credentials never enter a prompt
- Connector secrets are encrypted at rest and injected at call time. The model sees the tool signature, not the key.
- Sensitive actions wait for a person
- You choose which tools are gated and who can approve them. Approval happens in chat or in your team's messaging app, and both paths write the same audit record.
- Every run leaves a trace
- The plan, each tool call and its arguments, the approver and the timestamp are all recorded against the conversation.
- Organisations stay isolated
- Agents, connectors, documents and conversations are scoped to one organisation. Admins see their own and nothing else.
- A blocked request costs nothing
- Content guardrails run before the model is called, so a rejected request never leaves your deployment and never spends a token.
- Runs where you need it to
- Managed by us, or on your own infrastructure with Docker, PostgreSQL and Redis.
Pricing
Simple pricing, real capability
Two ways to get started, and a custom build for enterprises that need more.
Starter
SAR 499/mo
100M tokens / month
- Agents
- Skills
- Workflows
- Knowledgebase
- Connectors
- Artifacts
- Evals
Plus a one-time implementation fee, starting at SAR 5,000.
Pro
SAR 999/mo
500M tokens / month
- Agents
- Skills
- Workflows
- Knowledgebase
- Connectors
- Artifacts
- Evals
- Dedicated, fast customer support
Plus a one-time implementation fee, starting at SAR 5,000.
Custom
Custom
Usage sized to your rollout
- Everything in Pro
- Custom agent development
- Dedicated integration engineering
- SLA-backed support
Scoped with your team before rollout.
Talk to us
See it against your own systems
Pick a time below. Bring one workflow that crosses two of your systems and we will walk through what it looks like in Zamil.