Platform and integration
Canonical verbs
A canonical verb is one standard name for the same operation across every system that supports it, written domain.object.operation, such as service.ticket.create, so an agent learns an action once and can run it against any connected system.
What it is
Every vendor names the same operation differently: "create ticket", "add incident", "new service request", POST /tickets. A canonical verb gives that operation one name in three parts: a domain (service, finance, crm), an object (ticket, invoice, contact) and an operation (create, list, get, update, delete, and others).
The idea is related to resource-oriented API design. Google's AIP-121 says a small number of standard methods (Get, List, Create, Update, Delete) provide the semantics for most common operations. Canonical verbs apply that idea across vendors instead of within one API.
Why it matters
An agent that knows service.ticket.create can open a ticket in whichever help desk is connected without learning each vendor's vocabulary. It also lets a platform answer "which connected systems can do this?" and lets permissions be written once: allow service.ticket.create, not twenty-two vendor-specific names.
How it works
- Each action in each connector is tagged with a verb from a shared registry.
- An agent searches by verb, or by plain language, to find matching actions.
- A write must name the system it goes to. A read can be sent to every connected system that implements it at once.
- Each action keeps its vendor's own input schema; the verb names the operation, not the fields.
Example
crm.contact.list with no service named runs against every connected CRM that implements it and returns the results grouped by service. service.ticket.create with service: "halopsa" creates one ticket in HaloPSA.
Common pitfalls
- Same name, different inputs. A canonical verb does not mean identical arguments; read the action's schema before calling.
- Fanning out writes. Creating an invoice in every connected system at once is never what anyone wants.
- Verbs too broad to mean anything.
do.thinghelps nobody; a verb needs an object and an operation a person recognises.
How actionlake handles it
The actionlake registry defines 8,083 verbs. Connectors use 4,936 of them, and 3,353 are shared by two or more connectors. One name reaches many systems: service.ticket.create 54 connectors, finance.invoice.list 101, finance.invoice.create 74, crm.company.list 162 and crm.contact.list 177 (counted from our service files, 24 October 2026). find_actions filters by verb. A read verb with no service named fans out to every connected system that implements it, as one grouped run with one intent; writes are never fanned out. Keys can allow a verb rather than a vendor action. See the connectors.
Questions
What does service.ticket.create mean?
It is a canonical verb: the domain is service (service desk), the object is ticket and the operation is create. Any connector that can create a ticket carries it.
Do canonical verbs force a common data model?
No. The verb names the operation; each vendor's action keeps its own input schema. Read the schema of the action you are calling.
Can one canonical verb run against several systems?
A read can fan out to every connected system that implements it. A write must name one system.
Related terms
- Unified API vs MCPA unified API normalises many vendors in one category behind a single common schema for developers to code against, while MCP is a protocol that lets an AI model discover and call each system's own tools at run time; the two solve different problems and are often combined.
- AI agent permissionsAI agent permissions are the rules that decide which systems, actions and data an agent's credential may reach, and least privilege for agents means granting only the actions a task needs and enforcing that limit outside the model.
- Cross-system joinA cross-system join combines records from two or more separate systems that refer to the same real-world thing, such as one customer in an accounting ledger and a ticket desk, and returns a single combined answer.
- MCP toolAn MCP tool is a named function that an MCP server offers to AI models, described by a name, a plain-language description and a JSON Schema for its input, and invoked with a tools/call request.
- Action lakeAn action lake is a governed layer between AI agents and business systems where every action an agent may take is defined, every action it takes is authorised and priced, and every action it took is recorded, independent of the model, framework or vendor.
Sources
Connect an agent. Start free.
100,000 calls free every month. No card. Hard caps on every key.