Onboard a custom agent to the Recommended Actions Agent

The Recommended Actions Agent is a prioritization engine that:

Ingests raw action cards from multiple upstream agents such as Sales Opportunity Risk Agent, Data Enrichment Agent, and custom agents.
Scores each action on four dimensions — Urgency, Impact, Confidence, Effort (UICE framework, 0–10 each).
Ranks actions across all agents using a configurable priority formula.
Surfaces the top 5 actions in the Recommended Actions Agent carousel in Dynamics 365 Sales.

Your agent pushes actions into the Recommended Actions Agent pipeline via a Custom API. Recommended Actions Agent handles everything downstream: prioritization, deduplication, versioning, and UI rendering.

Integration architecture

The section describes the architecture and components involved in integrating a custom agent with the Recommended Actions Agent.

Processing Pipeline

Your agent detects an actionable insight (for example, a deal risk, a stalled deal, or a missing stakeholder).
Your agent calls msdyn_PushActionDataToRecommendedActionAgent Custom API to push the action.
The action is persisted in msdyn_rawactioncatalogue (input table).
For each action, the Scoring Engine (on the Copilot Service Fabric service RecommendedActionsAgentServices):
- Fetches entity signals from Dataverse.
- Fetches agent-specific prioritization data from the action catalogue.
- Calls LLM (via CAPI) to score on UICE dimensions.
- Applies floor/ceiling.
- Computes final priority score: U × 0.30 + I × 0.35 + C × 0.20 − E × 0.15 + PP × 0.05.
Scored action is inserted into msdyn_prioritizedactioncatalogue (output table).
Recommended Actions Agent Carousel fetches scored actions via msdyn_RAAgent_FetchScoredRecommendedActions and renders cards.

Key components

Component	Location	Description
Input table	`msdyn_rawactioncatalogue` (Dataverse)	Raw actions that upstream agents push
Output table	`msdyn_prioritizedactioncatalogue` (Dataverse)	Scored and ranked actions for the UI
Agent config	`msdyn_recommendedactionsourceagentconfig` (Dataverse)	Per-agent registration and configuration
Run tracking	`msdyn_recommendedactionrundetails` (Dataverse)	Processing state for each action
Push API	`msdyn_PushActionDataToRecommendedActionAgent` (Custom API)	Agent → Recommended Actions Agent action push
Scoring Engine	`RecommendedActionsAgentServices` (SF microservice)	DTF orchestration + LLM scoring
Fetch API	`msdyn_RAAgent_FetchScoredRecommendedActions` (Custom API)	UI → scored actions fetch
Solution	`NextBestActionAgent` (Dataverse solution)	Contains all entities, APIs, plugins, FCS

Prerequisites

Before onboarding, make sure that the following prerequisites are met:

NextBestActionAgent solution is deployed to the target org.
Feature Control Setting SalesService.NextBestAction:EnableNBASuggestions is set to true.
Feature Control Setting SalesService.NextBestAction:EnableRecommendedActionsAgent is set to true (or not explicitly false).
RecommendedActionsAgentServices SF microservice is deployed and healthy (CoPilot Service Fabric app).
Your agent has appropriate Dataverse security roles to call the msdyn_PushActionDataToRecommendedActionAgent API (requires prvWritemsdyn_rawactioncatalogue privilege).
You have a stable, unique SourceAgentId string for your agent (such as "DealClosingAgent", "SQAAgent").

Step 1: Register your agent (source agent config)

Add the agent to your organization through the agent creation process described in Add custom agents for recommended actions. During agent creation, select the option to add the agent to the recommended actions framework and provide the necessary information about the agent.
Register the agent in the msdyn_recommendedactionsourceagentconfig table. The Recommended Actions Agent uses this table to recognize and fetch actions from your agent. The key field that links your agent to the recommended actions framework is the msdyn_internalprioritizationinstruction field in the msdyn_syncactionexecutionstateapiconfig table. This field should match the SourceAgentId value you use when pushing actions via the API.

Open and update the table as needed to configure the agent's behavior in the recommended actions framework.

Request JSON schema:

{
  "agentName": "YourAgentName",
  "agentType": "CustomAgent",
  "isRecommendedActionAgentEnabled": true,
  "salesAgentProfileId": "<your-sales-agent-profile-guid-or-omit-to-auto-create>",
  "agentImpactMapping": "[\"Risk Mitigation\",\"Deal Velocity\"]",
  "internalPrioritizationInstruction": "{\"signals\":[...]}",
  "syncActionExecutionStateApiConfig": "{\"syncactionuistatusapiname\":\"your_SyncBackCustomApiName\"}",
  "description": "Brief description of your agent"
}

Important

Don't pass recommendedActionSourceAgentConfigId. The plugin resolves the target record by looking up an existing config via salesAgentProfileId. If no record exists, the plugin creates a new one. If agentType is "CustomAgent" and no salesAgentProfileId is provided, the plugin auto-creates and activates a new msdyn_salesagentprofile record.

Response: JSON with the upserted record ID and success status.

Full request fields (from UpsertRecommendationAgentConfigRequest.cs):

JSON Field	Type	Description
`agentName`	string	Maps to `msdyn_agentname` (max 850 chars). Required for new records.
`agentType`	string	Agent category. Use `"CustomAgent"` for non-Sales Opportunity Agent agents to auto-create a profile.
`isRecommendedActionAgentEnabled`	boolean	Maps to `msdyn_isrecommendedactionagentenabled`. Null = leave unchanged.
`salesAgentProfileId`	guid?	Links to `msdyn_salesagentprofile`. Used for record lookup on upsert.
`agentImpactMapping`	string	Flat JSON array of principle names (see below). Maps to `msdyn_agentimpactmapping`.
`internalPrioritizationInstruction`	string	JSON with `signals` array (see below). Maps to `msdyn_internalprioritizationinstruction`.
`syncActionExecutionStateApiConfig`	string	JSON object `{"syncactionuistatusapiname":"..."}`. Maps to `msdyn_syncactionexecutionstateapiconfig`.
`sourceAgentUniqueId`	string	Maps to `msdyn_sourceagentuniqueid`. Only written on create, ignored on update.
`description`	string	Maps to `msdyn_sourcedescription` (max 1000 chars).

Agent impact mapping

Maps your agent's actions to Recommended Actions Agent's prioritization principles. The scoring engine uses this mapping to weight your actions.

Format: A flat JSON array of strings — each string is a numeric optionset value. For example, if your agent focuses on Risk Reduction (4) and Deal Velocity (2), you can pass:

["4", "2"]

Available prioritization principles (from the global optionset msdyn_RecommendedActionAgent_SalesPrioritizationPrinciples):

Optionset Value	Label (use in `agentImpactMapping`)	Code Constant	Description
—	Org Custom Priorities	`PrincipleOrgCustomPriorities`	Admin-defined overrides — always highest weight
1	Revenue Generation	`PrincipleRevenueGeneration`	Actions that directly progress late-stage opportunities toward close
2	Deal Velocity	`PrincipleDealVelocity`	Unblock stalled deals and accelerate progression
3	Pipeline Generation	`PrinciplePipelineGeneration`	Generate new qualified pipeline (lead qualification, upsell/cross-sell)
4	Risk Reduction	`PrincipleRiskMitigation`	Address competitor threats, engagement silence, missing stakeholders
5	Customer Sentiment	`PrincipleCustomerSentiment`	Improve relationship health, respond to negative sentiment
6	Data Hygiene	`PrincipleDataHygiene`	CRM data quality — only when fields impact forecasting
7	Stakeholder Coverage	`PrincipleStakeholderCoverage`	Ensure decision-maker engagement, multi-threaded relationships

You can pass the numeric optionset value and the ResolvePrincipleDisplayValue() method resolves it:

Numeric → matched against OptionSetValue or Rank in the org's PrioritizationPrinciples config (from msdyn_salesagentconfigurationv2)
No match → raw value passed through as-is

Internal prioritization instruction

Agent-specific signal metadata that tells the scoring engine how to interpret your agent's prioritization data fields.

Format: A JSON object with a top-level signals array. Each signal is deserialized into AgentSignalInstructionConfig with these fields:

Field	Type	Description
`name`	string	Signal identifier - used as the key in the scoring prompt's Signal Reference section
`type`	string	Data type: `"string"`, `"number"`, `"boolean"`
`source`	string	Descriptive label for where the signal comes from. Not used for routing - `fetch_info.fetch_type` controls the actual fetch mechanism. Typically `"action_data"` for agent-pushed signals.
`dimension_influence`	`{dimension: strength}`	Which UICE dimensions this signal affects and how strongly. Keys: `"urgency"`, `"impact"`, `"confidence"`, `"effort"`. Strengths: `"strong"`, `"moderate"`, `"weak"`
`interpretation`	string	Natural-language description of what the signal means for scoring - injected into the LLM prompt
`reliability`	string	How reliable this signal is: `"high"`, `"medium"`, `"low"`
`required`	boolean	Whether the signal must be present for scoring
`fetch_info`	object	Controls where and how the signal value is retrieved at scoring time. See Fetch info section below.

Fetch info - `fetch_type` values

The fetch_type field inside fetch_info determines how the scoring engine retrieves the signal value. The following types are supported (from SignalFetchService.cs):

`fetch_type`	What it does	`crm_field` meaning
`action_data`	Reads the value from the per-action `msdyn_prioritizationdata` JSON. This is what custom agents should use.	JSON property key to look up in `msdyn_prioritizationdata`. Falls back to `name` if omitted.

Note

For custom agents: Use fetch_type: "action_data" for your agent-specific signals. Agent-level signals from internalPrioritizationInstruction are merged into the unified signal list via MergeAgentSignals() and all fetch types are dispatched. However, there's a critical limitation: the merge only copies FetchType and CrmField from your signal config - it doesn't copy SourceEntityName, TargetEntityName, SourceEntityField, SelectField, AggregationType, Derivation, FilterField, FilterValue, JsonPath, or ValueType.

What this means in practice:

action_data works perfectly - it only needs CrmField (the JSON key in msdyn_prioritizationdata)
entity_field silently fails - the engine filters by s.FetchInfo?.SourceEntityName == actionData.EntityTypeName, but SourceEntityName is null after merge, so the signal is skipped
linked_entity, aggregation, derived_insight also fail - they require TargetEntityName, SourceEntityField, etc. which aren't copied

If you need CRM-fetched signals (entity fields, linked entities, aggregations), configure these at the org level in msdyn_salesagentconfigurationv2 - they're shared across all agents. Your internalPrioritizationInstruction should only define action_data signals that read from your per-action msdyn_prioritizationdata JSON.

Examples per `fetch_type`

{
  "signals": [
    {
      "name": "risk_type",
      "type": "string",
      "source": "action_data",
      "dimension_influence": { "urgency": "moderate", "confidence": "weak" },
      "interpretation": "Risk category code assigned by the source agent (e.g. 8 = Missing BANT Info). Used for pre-filter rule matching and prompt context.",
      "reliability": "high",
      "required": false,
      "fetch_info": { "fetch_type": "action_data", "crm_field": "riskType" }
    },
    {
      "name": "risk_label",
      "type": "string",
      "source": "action_data",
      "dimension_influence": { "urgency": "weak", "confidence": "weak" },
      "interpretation": "Human-readable risk name from the source agent (e.g. 'Missing BANT Info', 'Stalled Pipeline'). Useful for prompt context and seller explanation.",
      "reliability": "high",
      "required": false,
      "fetch_info": { "fetch_type": "action_data", "crm_field": "risk" }
    }
  ]
}

How the three configs correlate (signal pipeline)

The three configuration fields form a three-layer signal pipeline that feeds the LLM scoring prompt:

Layer	Scope	Purpose	Example
`agentImpactMapping`	Agent config (per agent)	Defines which principles the agent serves.	`["1", "4"]`
`internalPrioritizationInstruction`	Agent config (per agent)	Explains how the LLM should interpret your signals.	`signals[].interpretation`, `dimension_influence`
`msdyn_prioritizationdata`	Per action (push API payload)	Carries actual signal values for each action.	`{"riskType":"8","risk":"Missing BANT Info"}`

Sales Opportunity Agent example for correlation

`internalPrioritizationInstruction` signal	`fetch_info.crm_field`	`msdyn_prioritizationdata` key	Example value
`risk_type`	`"riskType"`	`riskType`	`"8"`
`risk_label`	`"risk"`	`risk`	`"Missing BANT Info"`

So Sales Opportunity Agent's per-action msdyn_prioritizationdata JSON looks like:

{"riskType": "8", "risk": "Missing BANT Info"}

To onboard your own agent, follow the same pattern:

Define signals in internalPrioritizationInstruction with fetch_type: "action_data" and a crm_field key.
When pushing actions via msdyn_PushActionDataToRecommendedActionAgent, set msdyn_ActionPrioritizationData to a JSON object using those same keys.
If the keys don't match, the signal value is null and the LLM treats it as missing.

How dimension_influence affects scoring:

Each signal's dimension_influence is rendered into the prompt's Signal Reference like:

**risk_type** (string, source: agent): Risk category code... [urgency(moderate), confidence(weak)] reliability=high

The LLM uses these hints as guidance (not hard multipliers) to determine how strongly each signal affects each UICE dimension score.

Sync action execution state API config

Format: A JSON object (not a plain string) specifying the Custom API name that Recommended Actions Agent calls when a seller acts on your card, such as marking it as complete or irrelevant. This API sets the status of the action in the source custom agent.

{
  "syncactionuistatusapiname": "your_SyncBackCustomApiName"
}

Sales Opportunity Agent's actual value: {"syncactionuistatusapiname":"msdyn_SyncDealRiskActionFromNba"}

Step 2: Push actions via custom API

API: msdyn_PushActionDataToRecommendedActionAgent

Call this API each time your agent generates or updates an action for a target entity.

Request parameters

Parameter	Type	Required	Description
`msdyn_ActionId`	string	Yes	Your agent's unique identifier for this action. Used for deduplication and state sync. Must be deterministic (same action = same ID). Example format: `"DealRisk_{opportunityId}_{riskType}"`
`msdyn_SourceAgentId`	string	Yes	Your agent's identifier. Must match the `msdyn_agentname` in your agent config record. Example: `"DealClosingAgent"`
`msdyn_TargetEntityId`	uniqueidentifier (GUID)	Yes	The GUID of the target record (Opportunity, Lead) this action relates to
`msdyn_TargetEntityTypeName`	string	Yes	Logical name of the target entity. Example: `"opportunity"`, `"lead"`
`msdyn_ActionReason`	string	Yes	Reason why the action was generated. Used by the scoring engine for principle mapping. Example: `"Competitor Intel mentioned in recent email"`
`msdyn_ActionUIPayload`	string	No	JSON payload for card rendering (see Step 3). If omitted, Recommended Actions Agent can't display the card.
`msdyn_ActionPrioritizationData`	string	No	JSON with agent-specific data for scoring (see Step 4)
`msdyn_ActionCTA`	string	No	CTA type string. Example: `"Email"`, `"Review"`, `"Call"`
`msdyn_PrioritizationPrinciples`	string	No	JSON array of prioritization principles this specific action maps to (can override agent-level mapping)

Response Properties

Property	Type	Description
`msdyn_IsSuccess`	boolean	Whether the action was successfully persisted
`msdyn_Error`	string	Error message if `IsSuccess` is false

Example Call (C# Plugin)

var request = new OrganizationRequest("msdyn_PushActionDataToRecommendedActionAgent");
request["msdyn_ActionId"] = $"DealRisk_{opportunityId}_{riskType}";
request["msdyn_SourceAgentId"] = "DealClosingAgent";
request["msdyn_TargetEntityId"] = opportunityId;       // Guid
request["msdyn_TargetEntityTypeName"] = "opportunity";
request["msdyn_ActionReason"] = "Customer has not responded in 14 days, deal is at risk of stalling";
request["msdyn_ActionUIPayload"] = JsonConvert.SerializeObject(new
{
    version = 1.0,
    header = "Follow up with Contoso",
    description = "No customer response in 14 days. Deal may stall without re-engagement.",
    oncardClickActionType = "Navigate",
    oncardClickActionTypeParameters = $"{{\"etn\":\"opportunity\",\"id\":\"{opportunityId}\",\"pagetype\":\"entityrecord\"}}"
});
request["msdyn_ActionPrioritizationData"] = JsonConvert.SerializeObject(new[]
{
    new { signalName = "riskType", value = "14" },
    new { signalName = "risk", value = "low" }
});

var response = orgService.Execute(request);
bool success = (bool)response["msdyn_IsSuccess"];

Step 3: Action UI payload contract

The msdyn_ActionUIPayload field contains a JSON payload that controls how your action card appears in the Recommended Actions Agent carousel.

Schema

{
  "version": 1.0,
  "header": "Follow up with Contoso on pricing proposal",
  "description": "Stakeholder engagement has dropped. The customer expressed interest in the enterprise tier but hasn't responded to the last proposal sent 10 days ago.",
  "oncardClickActionType": "Navigate",
  "oncardClickActionTypeParameters": "{\"etn\":\"opportunity\",\"id\":\"<guid>\",\"pagetype\":\"entityrecord\"}",
  "onctaClickActionType": "Navigate",
  "onctaClickActionTypeParameters": "{\"etn\":\"opportunity\",\"id\":\"<guid>\",\"pagetype\":\"entityrecord\"}"
}

Field	Required	Description
`version`	Yes	Schema version. Use `1.0`.
`header`	Yes	Card header text — one-line action summary. Maps to `SuggestedActionItem.Title`.
`description`	Yes	Card body text — detailed explanation. Maps to `SuggestedActionItem.Description`.
`oncardClickActionType`	No	What happens when seller clicks the card. Currently only `"Navigate"` is supported.
`oncardClickActionTypeParameters`	No	JSON string with navigation parameters. Must include `etn` (entity type), `id` (GUID), `pagetype`.
`onctaClickActionType`	No	What happens when seller clicks the Go to record form option.
`onctaClickActionTypeParameters`	No	JSON string with CTA navigation parameters.

Step 4: Prioritization Data Contract

The msdyn_prioritizationdata field lets you pass agent-specific signals that influence how the UICE scoring engine prioritizes your action.

Schema

[
  { "signalName": "risk", "value": "low" },
  { "signalName": "riskType", "value": "4" }
]

How this data is used

The scoring engine reads these signals alongside entity-level signals (deal value, stage, competitors, and more).
Your msdyn_internalprioritizationinstruction from the agent config tells the LLM how to interpret each signal.
The scoring engine combines signals into the UICE scoring prompt.

Step 5: Handle action updates (versioning)

When your agent has updated data for a previously pushed action, create a new record with the same msdyn_ActionId by calling msdyn_PushActionDataToRecommendedActionAgent again.

How versioning works

The system creates a new row in msdyn_rawactioncatalogue with the same msdyn_actionid but a new msdyn_rawactioncatalogueid.
Recommended Actions Agent keeps showing the old version until it processes the new version.

Invalidating an action

If an action is no longer valid (for example, the risk was resolved):

Call msdyn_RAAgent_RemoveActionsV2 Custom API with the actionId.
This process marks all msdyn_rawactioncatalogue records for that action as Inactive.
The card disappears from the carousel.

Step 6: Bidirectional state sync

Recommended Actions Agent to your agent (seller acts in carousel).
When a seller marks an action as Done or Dismissed in the Recommended Actions Agent carousel:
1. Recommended Actions Agent updates the msdyn_actionuistatus in msdyn_prioritizedactioncatalogue.
2. Recommended Actions Agent reads your msdyn_syncactionexecutionstateapiconfig from the agent config.
3. Recommended Actions Agent calls your Custom API with:
  
  Parameter Type Description
  
  actionid GUID The action identifier
  
  state string "MarkedDone" or "Dismissed"
You must implement a Custom API that accepts these two parameters and updates the action state in your own data store.

Parameter	Type	Description
`actionid`	GUID	The action identifier
`state`	string	`"MarkedDone"` or `"Dismissed"`

Your agent to Recommended Actions Agent (seller acts in your UI).
When a seller acts on an action in your UI (for example, marks as mitigated on the custom agent page), sync that state to Recommended Actions Agent:

API: msdyn_SyncActionExecutionStateFromAgent

Parameter	Type	Required	Description
`msdyn_ActionId`	string	Yes	The action identifier (same as what you pushed)
`msdyn_ActionState`	integer	Yes	New state - values TBD (mapped to MarkAsDone/Dismissed)
`msdyn_TargetEntityId`	uniqueidentifier	Yes	Target entity GUID
`TargetEntityTypeName`	string	Yes	Target entity logical name
`msdyn_TrackingId`	string	No	Optional tracking/correlation ID

This action updates the state in the Recommended Actions Agent output table, hiding it from the carousel.

Step 7: Testing and validation

After you complete the configuration and implementation, test the end-to-end flow:

Verify agent config.

    GET [org-url]/api/data/v9.2/msdyn_recommendedactionsourceagentconfigs
    ?$filter=msdyn_agentname eq 'YourAgentName'
    &$select=msdyn_agentname,msdyn_agentimpactmapping,msdyn_internalprioritizationinstruction,msdyn_syncactionexecutionstateapiconfig

Push a test action.
Call msdyn_PushActionDataToRecommendedActionAgent with test data and verify:
- msdyn_IsSuccess = true
- A new record appears in msdyn_rawactioncatalogue with your action data
Trigger scoring.
Call msdyn_RAAgent_TriggerRecommendedActionsAgentOrchestration to trigger an on-demand scoring run (instead of waiting for the 4-hour timer).

Verify scored output.

    GET [org-url]/api/data/v9.2/msdyn_prioritizedactioncatalogues
    ?$filter=msdyn_actionid eq 'your-action-id'
    &$select=msdyn_actionid,msdyn_actionscore,msdyn_actionuipayload,msdyn_hascrossedceiling,msdyn_hascrossedfloor,msdyn_actionuistatus,msdyn_scoredetails

Verify:

msdyn_actionscore is populated with a value in the 0–10 range.
msdyn_hascrossedfloor is false (action is above floor and shows in carousel).
msdyn_actionuistatus is 1 (Active).
msdyn_scoredetails contains the LLM-generated explanation.

Verify carousel display.
Go to an Opportunity form in Dynamics 365 Sales and check the Suggested Actions section. Your scored action appears in the carousel.
Verify state sync.
1. Recommended Actions Agent → Your Agent: Dismiss an action in the carousel. Verify your sync-back API is called with state = "Dismissed".
2. Your Agent → Recommended Actions Agent: Mark an action in your UI. Verify the output table record has msdyn_actionuistatus updated.

End-to-end sales opportunity agent example (reference implementation)

Sales Opportunity Agent is the first agent onboarded to Recommended Actions Agent. Its integration serves as the reference implementation.

Agent configuration values (Sales Opportunity Agent)

Config Field	Sales Opportunity Agent Value (from `OraDefaults.cs`)
`msdyn_agentname`	`"SalesOpportunityAgent"`
`msdyn_agentimpactmapping`	`["DealRisk","Deal Velocity"]`
`msdyn_syncactionexecutionstateapiconfig`	`{"syncactionuistatusapiname":"msdyn_SyncDealRiskActionFromNba"}`
`msdyn_internalprioritizationinstruction`	See Sales Opportunity Agent's actual production value above

Action push (Sales Opportunity Agent to Recommended Actions Agent)

When Sales Opportunity Agent research completes and identifies deal risks, DealRiskToNBAService pushes each risk as a separate action:

Push Parameter	Sales Opportunity Agent Value
`msdyn_ActionId`	`"DealRisk_{opportunityId}_{riskType}"`
`msdyn_SourceAgentId`	`"DealRiskAgent"`
`msdyn_TargetEntityTypeName`	`"opportunity"`
`msdyn_ActionReason`	Risk description from research
`msdyn_ActionUIPayload`	Card with risk header + description
`msdyn_ActionPrioritizationData`	`{"riskType":"8","risk":"Missing BANT Info"}` (example)

State sync (Sales Opportunity Agent)

Sales Opportunity Agent → Recommended Actions Agent: When seller marks risk as done on research page → calls msdyn_SyncActionExecutionStateFromAgent
Recommended Actions Agent → Sales Opportunity Agent: When seller dismisses card in carousel → Recommended Actions Agent calls ora_UpdatedActionStateFromRAAgent (configured in agent config)

Feedback

Was this page helpful?

Last updated on 2026-06-04

Onboard a custom agent to the Recommended Actions Agent

Integration architecture

Processing Pipeline

Key components

Prerequisites

Step 1: Register your agent (source agent config)

Agent impact mapping

Internal prioritization instruction

Fetch info - fetch_type values

Examples per fetch_type

How the three configs correlate (signal pipeline)

Sync action execution state API config

Step 2: Push actions via custom API

Request parameters

Response Properties

Example Call (C# Plugin)

Step 3: Action UI payload contract

Schema

Step 4: Prioritization Data Contract

Schema

How this data is used

Step 5: Handle action updates (versioning)

How versioning works

Invalidating an action

Step 6: Bidirectional state sync

Step 7: Testing and validation

End-to-end sales opportunity agent example (reference implementation)

Agent configuration values (Sales Opportunity Agent)

Action push (Sales Opportunity Agent to Recommended Actions Agent)

State sync (Sales Opportunity Agent)

Related information

Feedback

Additional resources

Fetch info - `fetch_type` values

Examples per `fetch_type`