Oracle AI Agent Studio Connector Tools
A connector tool is one of the tool types you can give an agent in Oracle AI Agent Studio. It lets an agent reach a specific external client system through a pre-built connection rather than a hand-built REST call — most visibly the Slack and Microsoft Teams client connectors, which let people converse with a Fusion agent inside those messaging apps. Connectors sit alongside the other tool types (MCP tools, external REST tools, email tools, document tools) that agents use to sense and act on the world.
This page explains what a connector tool is, the connector types Oracle documents, how connectors are configured and secured, where human approval belongs, how to test them, and what to check at each Oracle release. Oracle AI Agent Studio is generally available since release 25C; the Slack and Teams channels that build on connectors arrived in 26A. It builds on the broader Oracle AI Agent Studio overview and the wider Oracle AI hub.
Last reviewed: 19 July 2026
Scope note — connectors vs channels vs MCP vs REST tools. These topics overlap and are easy to conflate, so this page draws the boundary explicitly:
- Connectors (this page) are the connector tool type — a pre-built link to a specific external client, including the Slack and Teams client connectors that underlie those channels.
- Channels are the user-facing surfaces where people reach an agent (Slack, Microsoft Teams). The connector is the plumbing; the channel is the experience. There is honest overlap here — see Oracle AI Agent Studio channels.
- MCP tools connect an agent to any external MCP-compliant server without a REST wrapper — see MCP tools in Agent Studio.
- External REST tools let an agent call an arbitrary third-party REST API. A connector targets a known client; a REST tool targets any endpoint you describe.
For the full catalogue of tool types and how they relate, see Oracle AI agent tools.
What Is a Connector Tool?
In Oracle AI Agent Studio, a tool is a capability you attach to an agent so it can do something beyond generating text — retrieve data, call a system, or send a message. A connector tool is the tool type that binds an agent to a specific external client system through a connection Oracle provides, rather than one you assemble from raw API descriptions.
The clearest confirmed examples are the Slack and Microsoft Teams client connectors. These let a user in Slack or Teams converse with a Fusion agent and receive its responses in that app. The connector handles the exchange with the messaging client; the channel configuration (26A) is what exposes the agent to those users. Because the connector underlies the channel, Oracle's readiness material describes them together — this page treats the connector as the tool type and defers the end-user surface to the channels page.
Connector tools live in the same catalogue as email tools (send email via the Fusion HCM Alert system, confirmed at 25C), document tools, MCP tools, and external REST tools. An agent can hold several tools at once and chooses between them within its instructions and topics. Keep the distinction in mind: a connector reaches a known client, an MCP tool reaches an MCP server, and a REST tool reaches an arbitrary API.
Availability note. Oracle AI Agent Studio is Generally available (release 25C). The Slack and Microsoft Teams connectors/channels are Generally available (release 26A). Beyond Slack and Teams, treat any specific connector product name as unconfirmed — describe connectors generically and confirm the current, licensed set of client connectors with Oracle for your environment.
Connector Types in Oracle AI Agent Studio
The table below summarises the connector-related tool types Oracle documents, with the availability labelling required for this vertical. Where a name is not an Oracle-confirmed product, the row describes the capability generically. Confirm the exact connectors licensed and enabled in your pod with Oracle.
| Connector | Purpose | Availability | Notes |
|---|---|---|---|
| Slack client connector | Let users converse with a Fusion agent inside Slack; deliver agent responses to the Slack client. | Generally available (release 26A) | Underlies the Slack channel; see the channels page for the user-facing surface. |
| Microsoft Teams client connector | Let users converse with a Fusion agent inside Microsoft Teams; deliver responses to the Teams client. | Generally available (release 26A) | Underlies the Teams channel; same connector/channel relationship as Slack. |
| Email tool | Allow an agent to send email via the Fusion HCM Alert system. | Generally available (release 25C) | Related outbound-messaging tool; not a client connector but often grouped with connectors. |
| Document tool | Attach documents an agent can search semantically as a knowledge source. | Confirmed capability | Feeds knowledge sources rather than a client system; distinct from client connectors. |
| Other client connectors | Connect agents to additional external client systems as Oracle expands the connector set. | Confirm current availability with Oracle | Described generically — do not assume a specific connector exists until confirmed for your release. |
| MCP tool (for contrast) | Connect to an external MCP-compliant server without a REST wrapper. | Generally available (release 26A) | A different tool type — see the MCP page; listed here to prevent conflation. |
| External REST tool (for contrast) | Call an arbitrary third-party REST API from an agent. | Enhanced in release 26A (incl. Require Human Approval) | A different tool type — targets any endpoint, not a known client. |
Configuring a Connector Tool
Configuration follows the same broad pattern as other Agent Studio tools: define the connector, supply the credentials that identify the target client, then attach the tool to one or more agents and describe when it should be used. Screens and field names change across releases, so treat the following as a generic checklist rather than a fixed screen path, and confirm the exact steps in Oracle's current documentation.
Connector configuration checklist
- 1. Identify the target client. Decide which external client the agent needs to reach (for example Slack or Microsoft Teams) and confirm that connector is available and licensed for your environment.
- 2. Register the connection. Provide the connection and credential details the connector requires so Agent Studio can authenticate to the client system.
- 3. Attach the tool to the agent. Add the connector tool to the relevant agent (or agent team) and reference it from the agent's topics and instructions.
- 4. Scope the behaviour. Use natural-language instructions to describe when the connector should be used and what it must not do, keeping actions within the agent's intended remit.
- 5. Set approval gates. Where the connector can trigger an action with a real-world effect, require human approval before execution (see below).
- 6. Validate in Agent Studio. Use the built-in debugging, validation and testing features to confirm the connector behaves as intended before you deploy the agent into a Fusion workflow.
For the Slack and Teams connectors specifically, most of the user-facing setup — registering the app in the messaging platform, mapping users, and exposing the agent — is described as channel configuration. See Oracle AI Agent Studio channels for that side of the work, and treat this page as the tool-type view.
Security and Authentication
A connector opens a path between a Fusion agent and an external client, so it inherits the usual integration questions: who is authenticated, what the agent may do through the connection, and what data crosses the boundary. Agents operate within Fusion's role-based permissions and policies, and a connector does not exempt them from that model.
- Credential handling. Store and manage the connection credentials through the mechanism Oracle provides; do not embed secrets in prompts or instructions.
- Least privilege. Give the connector — and the agent that uses it — only the access the task requires, and review that scope when roles or connectors change.
- Data exposure. Understand what data flows to the external client (for example, what an agent posts into a Slack channel) and confirm it is appropriate for that audience.
- Identity mapping. For channel connectors, confirm how a Slack or Teams user maps to a Fusion identity and what that user is authorised to see or do through the agent.
- Auditability. Ensure connector-triggered actions are traceable, so you can reconstruct what an agent did on whose behalf.
Governance and access-control practice for agents as a whole is a broader subject; this page stays on the connector-specific angle and links out for the operating model.
Human Approval Where Actions Execute
When a connector lets an agent do something with a real-world effect — post a message, notify a channel, trigger a downstream action — a human checkpoint belongs in front of that step. Oracle AI Agent Studio supports approval checkpoints in multi-agent orchestration, and the external REST tool type carries an explicit Require Human Approval option (enhanced at 26A) that pauses before an action executes. Apply the same principle to any connector-driven action that changes state or communicates outward.
Draw the line by consequence. A read-only exchange — an agent answering a question in Slack — carries less risk than an action taken on a user's behalf; the more irreversible or outward-facing the effect, the stronger the case for a person confirming it first. Confirm the current approval options in Oracle's documentation for the tool type you are using.
Testing Connector Tools
Connectors are where an agent meets an external system, so they deserve deliberate testing — functional, negative, security and release-regression. Oracle AI Agent Studio itself provides debugging, validation and testing features for agents; the matrix below outlines the connector-specific angles to cover.
| Test area | What to verify | Example |
|---|---|---|
| Connectivity | The connector authenticates and reaches the client system. | Agent delivers a response into the correct Slack workspace. |
| Functional path | The agent selects and uses the connector for the intended request. | A Teams user question routes to the agent and back. |
| Negative / failure | Bad credentials, downtime or a rejected message are handled gracefully. | Expired token surfaces a clear error, not a silent drop. |
| Approval gate | Actions that require approval genuinely pause for a human. | An outbound action waits until it is approved. |
| Security / access | Identity mapping and least-privilege scope hold. | An unauthorised user cannot trigger a restricted action. |
| Data exposure | Only appropriate data crosses to the external client. | Sensitive fields are not posted into a shared channel. |
| Release regression | Behaviour is unchanged after a quarterly update. | Re-run connector checks after a 26x update. |
Testing agents end to end — prompts, workflows, tools and approvals — is a broader discipline than connectors alone. For the wider testing method, see Oracle AI agent tools and the parent Agent Studio overview.
Release Considerations
Connector capability has moved quickly across recent Oracle releases, so availability matters as much as configuration:
- 25C — Oracle AI Agent Studio became generally available, with agent creation, tools, knowledge sources and the email tool (via the HCM Alert system).
- 26A — the Slack and Microsoft Teams channels/connectors arrived, alongside MCP tools, enhanced external REST tools (with Require Human Approval), and REST API access to agents.
- Ongoing — treat any connector beyond Slack and Teams as something to confirm against current Oracle readiness notes rather than assume, and re-test connectors after each quarterly update.
Because connector behaviour and the set of available connectors can shift release to release, a release-aware regression habit protects you from silent change. If you track Oracle's quarterly AI updates, tie connector re-testing to that cadence.
How SyntraFlow Helps
SyntraFlow is an Oracle Fusion test-automation and intelligence platform — it is not part of Oracle AI Agent Studio, and its own AI-driven testing engine is distinct from Oracle's agents. Where connector tools introduce new integration points into your Fusion estate, SyntraFlow can complement Oracle's built-in validation:
- SyntraFlow can be configured to include connector-driven agent behaviour in a regression suite that reruns on each Oracle quarterly update.
- SyntraFlow helps organizations assess where connector changes intersect with the Fusion processes those agents touch, so testing focuses on the highest-risk paths.
- SyntraFlow can connect release intelligence with test planning, flagging the quarters where connector and agent capabilities change so you can plan re-testing deliberately.
Configuration drift across environments is a common source of connector surprises. As a complementary control, Oracle Configuration Intelligence can be used to compare configuration between environments so a passing test reflects the setup you expect rather than a coincidental match. Treat these as capabilities to scope with our team, not guarantees.
Official Oracle References
These are the relevant Oracle sources for the facts on this page. Last reviewed: 19 July 2026.
Related Oracle AI Pages
Connectors are one tool type among several. Go deeper on adjacent topics:
Oracle AI Agent Studio ⭐
The parent Agent Studio overview.
Oracle AI Agent Tools →
The full catalogue of tool types.
Agent Studio Channels →
Slack & Teams as user channels.
MCP Tools →
Connecting to external MCP servers.
Oracle AI Hub →
The Oracle AI content hub.
Config Intelligence →
Environment configuration comparison.
Frequently Asked Questions
What is a connector tool in Oracle AI Agent Studio?
▼
It is a tool type you attach to an agent so it can reach a specific external client system through a pre-built connection. The clearest confirmed examples are the Slack and Microsoft Teams client connectors, which let users converse with a Fusion agent in those apps.
How is a connector different from the Slack or Teams channel?
▼
They overlap. The connector is the tool-type plumbing that links the agent to the messaging client; the channel is the user-facing surface where people actually reach the agent. The connector underlies the channel, so Oracle describes them together. For the end-user surface, see Oracle AI Agent Studio channels.
How is a connector different from an MCP tool or an external REST tool?
▼
A connector targets a known external client. An MCP tool connects to an external MCP-compliant server without a REST wrapper. An external REST tool calls an arbitrary third-party API you describe. All three are separate tool types in Agent Studio.
Which connectors are confirmed by Oracle?
▼
The Slack and Microsoft Teams client connectors are confirmed and generally available from release 26A. The email tool (via the HCM Alert system) is confirmed at 25C. Beyond those, confirm the current set of client connectors with Oracle for your environment rather than assuming a specific connector exists.
When is human approval needed for a connector action?
▼
Wherever a connector can execute an action with a real-world effect. Oracle AI Agent Studio supports approval checkpoints, and the external REST tool type has a Require Human Approval option (enhanced at 26A) that pauses before an action runs. Apply the same principle to any state-changing or outward-facing connector action.
How do you test connector tools?
▼
Cover connectivity, the functional path, negative/failure handling, approval gates, security and identity mapping, data exposure, and release regression. Oracle AI Agent Studio provides debugging, validation and testing features; layer connector-specific checks on top and re-run them after each quarterly update.
What are the security considerations for connectors?
▼
Manage credentials through Oracle's provided mechanism, apply least privilege, understand what data crosses to the external client, confirm how external users map to Fusion identities, and keep connector-triggered actions auditable. Agents operate within Fusion's role-based permissions, and connectors do not exempt them from that model.
Can SyntraFlow test Oracle agent connectors?
▼
SyntraFlow can be configured to include connector-driven agent behaviour in a regression suite and helps organizations assess where connector changes intersect with their Fusion processes. It is a complementary testing platform, distinct from Oracle's own AI Agent Studio validation — scope specifics with our team rather than assume coverage.
Assess Your Oracle AI Agent Connectors
Understand where Agent Studio connectors touch your Fusion estate, how to secure and approve their actions, and how to keep them tested through Oracle's quarterly updates. Book an assessment with SyntraFlow.