AI and Machine Learning Microsoft Foundry premium

Agent tool

Agent tool is a capability an agent can invoke to search, run code, query data, call APIs, or perform a configured action. In everyday Azure work, teams use it to let an agent go beyond text generation by grounding answers or carrying out controlled work. The useful evidence is tool type, configuration, authentication method, allowed operations, schema, endpoint, logs, and tool-call result. Treat the term as an operating handle, not trivia: know who owns it, which boundary it affects, what could break, and which Azure output proves the current state before a production decision.

Aliases
Foundry agent tool, AI agent tool, agent built-in tool, agent custom tool
Difficulty
intermediate
CLI mappings
4
Last verified
2026-05-09

Microsoft Learn

An agent tool is a capability that a Foundry agent can invoke to retrieve information, run code, call functions, search data, or use connected services. Tools extend an agent beyond model-only text generation, but each tool must be governed because it can expose data or perform actions.

Microsoft Learn: Agent tools overview for Foundry Agent Service2026-05-09

Technical context

In Azure architecture, Agent tool sits in the action and grounding layer that connects Foundry Agent Service agents to data sources and external capabilities. It works with built-in tools, custom tools, web search, code interpreter, Azure AI Search, Fabric, Logic Apps, APIs, identities, and application orchestration. The important distinction is whether the reader is inspecting configuration, runtime behavior, identity, billing, or observability evidence. A strong design records scope, owner, permissions, monitoring signal, and rollback path so the term can be checked consistently across development, test, and production environments.

Why it matters

Agent tool matters because it turns an Azure label into a decision point that operators can inspect, govern, and improve. Used well, it keeps work tied to evidence such as tool type, configuration, authentication method, allowed operations, schema, endpoint, logs, and tool-call result. Used poorly, a poorly scoped tool can leak data, perform unintended actions, create loops, or make the agent unreliable. The practical value is judgment: knowing which setting or record proves reality, which team owns the next action, and which failure mode to check first during a release, audit, incident, or cost review. Good entries make that decision path clear enough for production use.

Where you see it

Signals, screens, and Azure surfaces where this term usually becomes operational.

Signal 01

In the Azure portal or Microsoft Foundry/Azure service UI where agent, tool definition, connected resource, identity, function, search index, code interpreter, or custom API is configured.

Signal 02

In Azure CLI, SDK, REST, or ARM/Bicep evidence used to inspect the supporting resources.

Signal 03

In governance workbooks, incident reviews, architecture diagrams, and runbooks where ownership and state are documented.

When this becomes relevant

Specific situations where this term helps solve real Azure design, operations, migration, security, reliability, cost, or governance problems.

  • Add search retrieval to an agent
  • Call a business API from an agent
  • Run code or data analysis through a tool
  • Govern agent access to connected services

Real-world case studies

Different enterprise-style examples that show the term being used to hit measurable objectives.

Case study 01

Agent tool in action

Scenario, objectives, solution, measured impact, and takeaway.

Scenario

Wingtip Freight, a logistics company, wanted a customer-support agent to check shipment status and open escalation cases without giving it broad system access.

Business/Technical Objectives
  • Let the agent retrieve live shipment status from an approved API
  • Require human approval before opening escalation cases
  • Log every tool call and result for support review
  • Reduce routine shipment-status tickets by 35 percent
Solution Using Agent tool

The engineering team configured an Agent tool that called a shipment-status API with a managed identity and a narrow schema. A second custom tool drafted escalation cases but returned them for human approval instead of submitting automatically. Tool inputs were validated against shipment ID format rules, and outputs were filtered to remove internal routing notes. Agent run logs captured tool calls, latency, and errors, while support supervisors reviewed failed tool attempts weekly. The agent instructions explicitly stated when to call each tool and when to escalate to a human.

The team also added owner tags, a rollback note, and a validation checklist so support, security, and finance reviewers could repeat the pattern without rebuilding the design from memory.

Results & Business Impact
  • Routine shipment-status tickets dropped 38 percent
  • No unauthorized case creation occurred during the pilot
  • Tool-call error rate fell from 11 percent to 3 percent after schema tuning
  • Average customer response time improved by 42 percent
Key Takeaway for Glossary Readers

Agent tools are powerful because they let agents act, but production value depends on tight scopes, validation, and reviewable tool results.

Case study 02

Agent tool in action

Scenario, objectives, solution, measured impact, and takeaway.

Scenario

Maple Grove Bank, a regional bank, needed an internal assistant to answer branch-policy questions and calculate fee examples without accessing customer accounts.

Business/Technical Objectives
  • Ground policy answers in approved internal documents
  • Run safe calculations through a code tool
  • Block access to customer-account systems
  • Provide citations and tool evidence for compliance
Solution Using Agent tool

The AI team enabled two Agent tools: a knowledge-search tool over approved branch-policy documents and a code interpreter tool for non-customer fee examples. The assistant could not call account systems or retrieve customer records. Tool configuration was reviewed with risk officers, and every tool result was logged with the agent run. The team tested prompt attempts that asked for unauthorized customer lookups and confirmed the assistant declined them. Branch employees received policy answers with citations and clearly labeled calculation examples.

The team also added owner tags, a rollback note, and a validation checklist so support, security, and finance reviewers could repeat the pattern without rebuilding the design from memory.

Results & Business Impact
  • Policy lookup time fell by 49 percent
  • Compliance sampling found citations in 96 percent of reviewed answers
  • Unauthorized account-access tests were rejected consistently
  • Branch training tickets dropped by 27 percent
Key Takeaway for Glossary Readers

Agent tools should be designed around allowed tasks, not around everything the agent might possibly want to do. The release team also kept the evidence reusable for the next review.

Case study 03

Agent tool in action

Scenario, objectives, solution, measured impact, and takeaway.

Scenario

Apex Clinical Trials, a life sciences services firm, wanted a protocol assistant that could search trial documents and create draft task lists for study coordinators.

Business/Technical Objectives
  • Search protocol documents without exposing unrelated studies
  • Generate draft coordinator task lists for review
  • Track tool-call evidence for regulated workflows
  • Reduce manual protocol lookup time by 30 percent
Solution Using Agent tool

The architecture team configured an Azure AI Search Agent tool scoped to each study’s approved index and a custom planning tool that generated draft tasks from retrieved protocol sections. Tool authentication used project-specific identities, and coordinators had to review task lists before adding them to the study system. The team logged which documents were retrieved, which tool produced the draft, and which user approved it. A disabled-by-default emergency switch let operators remove the planning tool if output quality dropped.

The team also added owner tags, a rollback note, and a validation checklist so support, security, and finance reviewers could repeat the pattern without rebuilding the design from memory.

Results & Business Impact
  • Protocol lookup time decreased 34 percent
  • Draft task-list preparation fell from 45 minutes to 18 minutes
  • No cross-study retrieval was found in validation tests
  • Tool evidence satisfied the internal quality audit
Key Takeaway for Glossary Readers

Agent tools let AI assistants touch real workflows, so scope, approval, and auditability matter as much as model quality.

Why use Azure CLI for this?

Azure CLI is useful for Agent tool because it turns portal knowledge into repeatable evidence. Tool configuration is often SDK/REST based; Azure CLI helps inspect backing resources, identities, secrets, and network access. Use CLI when you need inventory, comparison between environments, release notes, audit proof, or a safe pre-change check. Prefer read-only commands first, save structured output when possible, and treat mutating commands as change-controlled work with subscription, resource group, identity, and rollback details verified before execution.

CLI use cases

  • Inventory the Azure resources or records related to Agent tool and confirm the expected scope.
  • Inspect tool type, configuration, authentication method, allowed operations, schema, endpoint, logs, and tool-call result before a release, audit, incident review, or cost discussion.
  • Compare development, test, and production settings so drift is visible before users are affected.
  • Export structured evidence for tickets, runbooks, compliance reviews, or post-incident timelines.

Before you run CLI

  • Confirm the signed-in tenant, subscription, resource group, and target resource name before trusting output.
  • Check whether the command is read-only, mutating, credential-revealing, or potentially destructive.
  • Use the least-privileged identity that can inspect the resource and avoid pasting secrets into shared channels.
  • Decide the output format first, usually table for humans and JSON for automation or saved evidence.
  • Know the rollback or revoke path before running any command that changes state or permissions.

What output tells you

  • The output should identify the current Azure scope and show whether Agent tool is configured, active, enabled, or producing evidence.
  • Status, timestamps, IDs, names, and related resource references help connect Agent tool to a real owner and workload.
  • Empty output is still evidence: it may mean the feature is disabled, the wrong scope was queried, or the caller lacks permission.
  • Differences between environments usually point to drift, incomplete deployment, stale configuration, or an undocumented exception.

Mapped Azure CLI commands

Agent tool operator commands

direct
az search service show --name <search-service> --resource-group <resource-group>
az search servicediscoverAI and Machine Learning
az keyvault show --name <key-vault> --resource-group <resource-group>
az keyvaultdiscoverAI and Machine Learning
az keyvault secret list --vault-name <key-vault>
az keyvault secretdiscoverAI and Machine Learning
az cognitiveservices account deployment list --name <account> --resource-group <resource-group>
az cognitiveservices account deploymentdiscoverAI and Machine Learning

Architecture context

In Azure architecture, Agent tool sits in the action and grounding layer that connects Foundry Agent Service agents to data sources and external capabilities. It works with built-in tools, custom tools, web search, code interpreter, Azure AI Search, Fabric, Logic Apps, APIs, identities, and application orchestration. The important distinction is whether the reader is inspecting configuration, runtime behavior, identity, billing, or observability evidence. A strong design records scope, owner, permissions, monitoring signal, and rollback path so the term can be checked consistently across development, test, and production environments.

Security

Security for Agent tool starts with knowing the access boundary it creates or exposes. Review authentication, least-privilege tool scopes, input validation, output filtering, audit logs, and controls over side-effecting actions before trusting the configuration in production. Least privilege, source verification, and clear ownership matter because a small Azure setting can change who can read data, trigger actions, approve permissions, or serve user traffic. Security teams should capture evidence in tickets or runbooks without leaking secrets, tokens, sensitive payloads, or customer data. When possible, pair the term with Microsoft Entra roles, managed identities, policy, logging, and alerting so changes are visible, reviewable, and reversible.

Cost

Cost impact for Agent tool may be direct or indirect, but it should still be explicit. The main cost consideration is that tools can trigger paid search, code execution, external APIs, workflow runs, or storage reads beyond model token cost. Even when the term is not a billing meter, it can influence the services, retries, alerts, storage, model tokens, compute, or operations effort consumed around it. FinOps review should ask whether the setting is needed, who pays for it, how long evidence is retained, and whether tags, budgets, exports, or Advisor data make the spend explainable. Review the pattern whenever environments are cloned, scaled, or retired.

Reliability

Reliability depends on how Agent tool behaves during failure, scale, retries, and change windows. The main reliability concern is well-designed tools return predictable outputs, handle failures, expose status, and keep agent runs from stalling or hallucinating. Operators should know whether the term affects runtime traffic, orchestration state, alert delivery, recovery evidence, or only management-plane reporting. Before changing it, confirm the rollback path, expected health signal, blast radius, and dependency map. During incidents, use the term to narrow the question: what changed, what is active, what failed, and what evidence proves that the system can safely continue or recover? Keep that evidence close to the change record.

Performance

Performance impact for Agent tool depends on where it sits in the workload path. The main performance factor is tool latency, result size, retries, and parallelism strongly affect the total agent response time. Some terms do not speed the application directly, but they improve operational performance by reducing investigation time, noisy processing, or manual triage. Review latency, throughput, queue depth, query shape, token usage, retry behavior, and data volume where they apply. The best test is practical: can the team prove the term improves user experience, deployment speed, incident response, or processing efficiency without hiding a new bottleneck? Measure before and after; assumptions are not evidence.

Operations

Operationally, Agent tool should be part of a repeatable runbook, not a portal-only memory. Teams need a standard way of registering tools, testing schemas, monitoring calls, rotating credentials, and disabling tools during incidents. The runbook should name the Azure scope, owner, required role, normal state, change procedure, evidence to collect, and escalation path. Good operators also record why a value exists, not just what it is. That context prevents accidental cleanup, noisy alerts, unsafe reruns, stale dashboards, and confusing handoffs between platform, application, data, security, and finance teams. It also makes later reviews faster and less political. This keeps reviews repeatable when pressure is high.

Common mistakes

  • Treating Agent tool as a label instead of checking the Azure output that proves its current state.
  • Using the wrong tenant, subscription, project, database, or resource group and then trusting misleading results.
  • Saving sensitive keys, payloads, user data, or permission details in screenshots instead of sanitized evidence.
  • Changing production configuration without documenting the owner, rollback path, alert impact, and expected verification signal.