Softlander Book a call
Agent-to-agent interoperability

Agent capability discovery design

Design how agents find relevant capabilities without overtrusting descriptions

A coordinator chooses tools or peer agents dynamically. The decision is how to design how agents find relevant capabilities without overtrusting descriptions. The work begins with your actual workflow and looks for the evidence that separates intended behaviour from a consequential exception. It is particularly relevant to agent platform architects and discovery owners. A useful outcome is an explicit account of what was examined, what remains uncertain, and which decision the evidence can support.

Book a 30 minute readiness callBrowse this capability group

The interaction contracts relationship

  1. Starting evidenceCapability metadata

    Establish the basis for declared function.

  2. Decision boundaryVerified compatibility

    Check task and interface fit.

  3. Decision artefactCapability discovery model

    Retain the scoped finding and its unresolved dependencies.

Proposed working model. Untested peers require their own compatibility evidence.

Decision

Frame the decision around declared function

Questions and the evidence they need

Criterion / evidenceDeclared function
Inspect capability metadata for the relevant constraint or observation.
Criterion / evidenceVerified compatibility
Inspect registries for the relevant constraint or observation.
Criterion / evidencePermitted authority
Inspect eligibility rules for the relevant constraint or observation.
Read each criterion against the available evidence before deciding that additional work is needed.

The decision is how to design how agents find relevant capabilities without overtrusting descriptions. This is relevant to agent platform architects and discovery owners when a coordinator chooses tools or peer agents dynamically. Name the affected workflow, the consequence of a wrong decision, and the owner who can act on the result. The initial discussion should establish which evidence is already available and which uncertainty is material enough to investigate.

The specific failure hypothesis is that a fluent capability description leads to selection of an agent lacking the required permission or evidence. It remains a hypothesis until examined in the deployment. Existing internal tests or operating records may already resolve it; identify the uncovered question before commissioning additional work.

Mechanism

Advertised capability is only one selection input

Relationships within interaction contracts

Source or conditionDeclared function
Read what the peer says it can do.
Crossing or choiceVerified compatibility
Check task and interface fit.
Consequence or decisionPermitted authority
Apply eligibility before assigning the work.
Compare authority and state at the receiving boundary. Each relationship needs evidence in the actual scope.

What this specialty examines

Capability discovery needs separate fields for advertised function, verified compatibility and allowed authority. An agent may describe a task broadly while supporting only a narrow operation. Selection should consider the intended input, output, environment and constraints.

The review should separate advertised function, verified compatibility, permitted authority and selection policy. Identify the normal route, relevant exceptions and the authoritative record of the outcome. Inspect the relevant business-system record, enforcement event or process evidence alongside any reported status. Keep incomplete cases visible, with an explanation of what their absence prevents the owner from concluding.

The tradeoff to make explicit

For this decision, dynamic discovery improves flexibility while advertised capability requires bounded verification. Compare the available design or operating choices against the same intended outcome. If one option removes a required control or changes the task boundary, describe that difference before treating its apparent simplicity, speed or lower effort as an improvement. The decision should expose the tradeoff instead of burying it in an aggregate score.

Preparation

Connect the working inputs to the boundary being examined

Evidence packet for declared function

Illustrative working format

BasisCapability metadata
Establish permitted access, provenance, and which task conditions it represents.
ContextRegistries
Check its relationship to the relevant workflow and expose missing context.
ObservationEligibility rules
Confirm the applicable revision and who resolves disputed interpretation.
DependencyObserved task requirements
Confirm the applicable revision and who resolves disputed interpretation.
Illustrative preparation record. Availability, revision, and handling need confirmation.

Start with capability metadata, registries, eligibility rules and observed task requirements. Identify the owner, revision and operating context of each source. Sanitised examples may support initial design, but differences from actual conditions should remain explicit. Where direct access is unsuitable, a customer-controlled evidence or execution route may be considered, with agreed handling and retention.

The scope also needs the integration owner, agreed expectations for a representative interaction case, and the authority to inspect the relevant boundary. Specify what evidence would distinguish the failure in which a fluent capability description leads to selection of an agent lacking the required permission or evidence from acceptable behaviour. If stakeholders disagree about intended behaviour, resolve the interpretation before classifying the observed difference as a defect.

How the work proceeds

Carry declared function through to an accountable finding

From the scoped material to a decision record

  1. Establish conditionsCapability metadata

    Agree the task meaning at the sending boundary.

  2. Examine the relationshipVerified compatibility

    Apply the scoped check to permitted authority.

  3. Retain the findingCapability discovery model

    Record the evidence, open dependencies, and the owner’s next decision.

Proposed sequence; resolve interpretation or access gaps before relying on the next step.

Fix the interaction contract and versions

Assemble capability metadata, registries, eligibility rules and observed task requirements with agent platform architects and discovery owners. Identify the participating implementations, interface revisions, authority assumptions and intended business result. Prepare messages and task-state examples from both sides, including an interrupted exchange and a disputed or incomplete outcome. The preparation output is a paired contract fixture showing what each participant expects and which differences would prevent the workflow from completing correctly.

Run and compare both sides of the exchange

Define metadata ownership and validation before dynamic selection. Test stale descriptions, unsupported tasks and unavailable peers. Record how the coordinator detects a mismatch and falls back without assuming that another persuasive description supplies missing evidence.

Choose the contract or integration change

Summarise the comparison in a capability discovery model with metadata validation and selection criteria. Classify each mismatch as interface validation, business meaning, identity, task state or operating behaviour. Decide whether to change a participant, add an explicit adapter rule, narrow supported combinations or defer the connection. Validate the affected pair again and preserve limitations that translation cannot resolve, especially around cancellation, authority and partial completion.

Indicative artefact

Make the capability discovery model usable by its owner

Capability discovery model

Illustrative working format

Annotated fieldCapability discovery model
Identify the assessed workflow, revision, and decision boundary.
Annotated fieldMetadata validation
Link supporting and contrary observations to inspectable evidence.
Annotated fieldSelection criteria
Define the required observation and who resolves disagreement about acceptance.
Illustrative artefact structure, not a completed engagement or an observed result.

A useful decision artefact is a capability discovery model with metadata validation and selection criteria. Supporting material may include selected cases, configuration or process records, evidence references, a disagreement log and prioritised actions. Retain the case selection rationale and material disagreements alongside the result; otherwise the result may be mistaken for complete coverage.

Illustrative scenario

Examine permitted authority through a constructed exception

A constructed case and its decision paths

Illustrative scenario

  1. SituationDeclared function

    A discoverable agent advertises a task it lacks permission to execute.

  2. Question to examineVerified compatibility

    Check task eligibility and granted authority after capability discovery.

  3. Conditional next actionPermitted authority

    Exclude the peer until both compatibility and permission are supported.

Fictional example. No executed test, customer engagement, or measured outcome is implied.

Consider a fictional situation: an agent advertises invoice handling but only supports document extraction, not approval or payment actions. The scenario is illustrative, with no client engagement, executed test or measured improvement implied.

For invoice handling, distinguish extraction from approval and payment. The coordinator can select the extraction capability for document parsing while retaining approval elsewhere. A discovery label should not promote a limited function into authority over the whole process.

Scope and next action

Keep the conclusion within named implementations and revisions

Confirm fit before the next commitment

Included boundaryDeclared function
This review is limited to named implementations and revisions.
PrerequisiteCapability metadata
The relevant owner must confirm access and the intended operating conditions.
Claim limitCapability discovery model
Untested peers require their own compatibility evidence.
Bring the triggering change and existing evidence; resolve the dependency that could alter the decision.

Compatibility is assessed for named implementations, versions, features and workflows. It does not establish universal conformance, vendor endorsement or interoperability with untested peers. Any engagement depends on an agreed scope, qualified team, authorised access and confirmed availability. Bring the affected workflow, its current configuration or process, the triggering change and the evidence already available. The immediate scope question is prompted by the situation in which a coordinator chooses tools or peer agents dynamically; confirm which missing observation would change the owner’s next action.

Ways to engage

Related engagement scopes

AI Agent Protocol Interoperability Lab

Find where connected agents disagree. Exercise message contracts, identity propagation and failure recovery across your selected integrations.

Read about this offer

Continue exploring

Related insights and resources

Evaluation & assurance6 MIN READ

How to design a joint AI evaluation lab

A joint lab should be a governed way to create decision evidence across organisations, with shared methods but explicit ownership of systems, data and conclusions.

Agent security6 MIN READ

Agent protocols are creating a new control surface

Protocols make tools and agents interoperable, but they also standardise discovery, delegation and context exchange. Those interfaces need explicit trust and policy controls.

Agent security6 MIN READ

A practical security review for an MCP server

Review an MCP server as a privileged protocol adapter: verify tool semantics, OAuth boundaries, token audience, downstream delegation, input handling, isolation, logging and consent.

A little more clarity

Questions to consider.

Start with the questions that shape a better decision.

Can a capability description serve as evidence that an agent can perform the task?

No. Descriptions help discovery but need validation against the intended task, supported interface and permitted authority. A broad label should not be treated as proof of every implied operation.

Which metadata supports capability selection?

Start with capability metadata, registries, eligibility rules and observed task requirements. Record their revision, owner and relationship to the intended workflow. Missing context should be identified before interpreting outcomes; a sanitised fixture is useful only when its relevant differences from the operating case are understood.

How should advertised and verified functions be compared?

Separate advertised function, verified compatibility, permitted authority and selection policy. Compare both participants' accepted inputs, task states and results under a named interface revision. Include interrupted or delayed interactions. Parser acceptance should not be treated as agreement about business meaning, authority or completion.

What belongs in a capability discovery model?

The proposed output is a capability discovery model with metadata validation and selection criteria. Include the selected conditions, supporting observations, unresolved differences and next decision. A reviewer should be able to follow each important conclusion back to its evidence and understand the limits without depending on an undocumented explanation from the author.

When should an agent reject a discovered candidate?

Reject a candidate when its verified function, permissions, interface or availability does not meet the task. A persuasive capability description should not resolve missing evidence or expand the mandate.

What if existing evidence already answers the question?

Review its scope, provenance and configuration before proposing new work. Existing work from agent platform architects and discovery owners may already be sufficient. In that case, capture the decision, unresolved limitations and the change that would require another review. Additional assessment should address a real gap.

What would make the findings inconclusive?

Missing capability metadata, registries, eligibility rules and observed task requirements can prevent a reliable interpretation, depending on the case. Record the specific missing observation and its implication. The owner can supply evidence, narrow the claim or defer the dependent decision; an incomplete case should not silently be counted as a pass.

Start with the decision you need to make.

Tell us about the agent workflow you want reviewed, and where you need greater clarity. Scope and availability are confirmed together.

Book a 30 minute readiness call Get the agent release checklist

Connected disciplines

Delegation contract design

Make delegated work precise enough to accept, reject or recover. The review connects task requests, input schemas, authority limits and acceptance criteria to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.

Agent message-schema governance

Control message changes without breaking agent interactions. The review connects schema registries, consumers, version policies and message examples to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.

Protocol authentication review

Verify who is connecting through an agent protocol. The review connects issuer configuration, audience expectations, peer identity rules and transport setup to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.