REDA AI Monsoon · v1.16000 13 Sep 2026

Your Salesforce agents in a connected AI assistant

This main release brings together the changes since 1.15000, including patches 1.15001–1.15009. It extends access to your configured agents while improving action handling, administration and credit visibility. Upgrade the package before using the new settings; obtain the installation details through your REDA account executive or support@reda.one.

Connect selected agents through MCP

MCP connects an approved AI assistant to selected active Internal agents. Each person signs in with their Salesforce identity. The host receives agent entry points; individual REDA tool and prompt definitions are not exposed as separate MCP tools. Configure a Salesforce Site, an External Client App, both REDA AI User and REDA AI MCP User assignments, and agent opt-in. See the administrator setup guide and user guide. Interaction data can reach the connected host and configured AI providers as well as Salesforce. Review the intended data and agreements; this is not a promise that data never leaves Salesforce.

Related instructions

• About REDA AI
• App Access & Permissions
• Understanding Agents
• Internal Channel
• Data Access & Permissions Considerations
• Data Security in REDA AI
• Set up MCP access for REDA AI
• Use REDA AI with MCP

Understand the Economy credit change

Stable instructions are reused across conversations to improve cache reuse. When Enhanced Caching is enabled, its longer cache duration starts with the next cold cache fill; it does not extend an entry that is already warm. Record context stays fresh as the conversation changes.

Economy ordinary usage now applies a 0.40 multiplier to the base credit rate. For unchanged token quantities and base rates, an example costing 100 base credits costs 40 Economy credits. At the previous 0.10 multiplier it cost 10: the new debit is four times the previous Economy debit, and 60% below base. This changes credit consumption, not the subscription price. File pre-extraction is charged separately at full rate. A provider fallback can use a different effective tier; inspect the actual transaction rather than applying 0.40 to every mixed operation. New credit transactions record Processing Tier; historical rows are not backfilled. Economy still requires consent and compatible licence/provider configuration; a missing consent is refused, not silently processed at another tier.

Related instructions

• How Credits Work
• Credit Usage Optimisation
• Configuring Agent Settings
• Conversation Analytics

Review conversation continuity

A blank Conversation Window (Hours) now uses the organisation default, initially 30 minutes. A positive value sets the agent's window; 0 disables automatic resume. Matching remains channel-specific: API uses agent and record context; authenticated Internal/MCP uses agent, user and the relevant record context. Community uses the shared window setting instead of the external-session timeout. Home, app-page and utility-bar chat can resume eligible conversations. Explicit conversation identifiers remain subject to their route's access checks. Review agents that relied on blank meaning a new conversation, and use 0 where that behaviour is required. Packaged permission fixes also address Internal and Community conversation access.

Related instructions

• Configuring Agent Settings
• Internal Channel
• Community Agent
• Viewing Conversations & Transcripts
• Community API Reference

Clearer email, record and scheduled actions

Task and event tools can accept additional field values, including permitted Contact/Lead associations. Agent-sent emails use the configured branding on supported server-executed paths; test a draft or an internal recipient after merging your tool definitions.

Email actions distinguish a Contact or Lead recipient from the related business record, support To/Cc lists, and check the supported recipient opt-out cases. Contact/Lead values supplied as related records are handled as recipients instead of invalid activity links. Record confirmations surface fields that were not applied, rather than silently implying success. Plan steps are reconciled one-to-one; relative scheduled times retain UTC correctly. Disabled agents and disallowed nested actions are checked before confirmed execution. File-link confirmation reuses the supported access-aware path. New Pending or rescheduled time-based actions require a runnable scheduled processor; email-reply tracking and ordinary history/status updates remain valid. Delayed jobs run as their scheduler owner, not by impersonating the requestor. Host guidance reduces duplicate use of another email/calendar connector; it cannot guarantee the host's behaviour.

Related instructions

• Action Confirmations
• Built-in Tools Reference
• Scheduled Actions
• Card-List Rendering
• Email Channel — Outbound
• Branded Email

Configure reusable business instructions

Exclusive Prompts let administrators assign reusable instructions to selected agents. Review the Exclusive Prompt flag and Summary on Custom Prompt, then the assignment's Agent Definition, Custom Prompt, Active and Load Mode. Use the intended load mode; On Demand needs the context-prompt tool in the agent's loadout. Record-type brackets use shared scope rules, including in preview and prompt administration. Prompt and tool ordering is deterministic, and runtime preview can use an actual record ID. Updated language instructions follow the message on API/external channels; user placeholders apply only where the authenticated user context is intended. Preserve custom prompts when merging the current defaults.

Related instructions

• Custom Prompts Per Agent
• Creating & Editing Prompts
• Understanding Prompts
• Agent Runtime Preview
• Understanding Context Fields
• Multi-Language Agents
• Prompt Template Gallery

Automation and reliability

The per-agent hourly rate limit also applies during the final UTC hour of the day. External and Community APIs use a stable credit-limit error code so the intended HTTP 429 response no longer depends on the translated message.

Bulk Flow invocations can return deferred request IDs instead of immediate answers. Review downstream Flow elements that expect response, credits or action-result output in the same transaction. Automated action handling carries the configured channel and execution mode, and retains actual outcomes in conversation context; this does not guarantee an extra narrated turn. Explicit datetime offsets are preserved in supported event, scheduled-action, field-write and CSV paths. Offset-free values retain the running user's interpretation. Routine System-mode processing no longer raises warning incidents merely because it uses that mode; ERROR-level notifications and incident logs remain. File extraction respects the configured output limit, semantic status colours work across the supported Salesforce themes, and packaged tests avoid static Account record-type references that blocked installation in orgs without those record types.

Related instructions

• Flow Integration Reference
• Async Agent Processing
• AI Incident Logs
• Document Processing
• External Widget API Reference

Client steps to start using 1.16000

Complete this checklist for your installed baseline. It includes patches 1.15001–1.15009 as well as main release 1.16000. Features are optional; complete their setup before enabling them. The field table lists all 13 packaged fields added since main release 1.15000: nine in patches and four in 1.16000.

1. Record the baseline and preserve configuration

In Setup → Installed Packages, record the current REDA AI version. Test the upgrade with your own configuration in a suitable sandbox. Back up agent settings, custom prompts, tool custom metadata, layouts, permission sets and pending schedules. If upgrading from before 1.15000, also complete the 1.15000 upgrade steps. Install 1.16000 using the details supplied by REDA, then verify the version again.

2. Place the fields where your team works

For each object in the field table, open Setup → Object Manager → object → Page Layouts. Use Page Layout Assignment to identify every layout used by the relevant profiles and record types. Edit each applicable layout, add the listed fields to the specified section, and save. A package upgrade does not add fields to a subscriber’s customised layout. If the Lightning record page uses Dynamic Forms, also open Edit Page, add the fields to its field sections, review visibility rules and activate the appropriate page assignment. A page-layout change alone is insufficient for Dynamic Forms.

On the Agent Definition layout, add the Exclusive Prompts related list so administrators can create assignments. Include the prompt, Active and Load Mode columns. Make the new Exclusive Prompt object and its layout accessible through the app or an approved related-list workflow. Keep Via MCP, credit Processing Tier and Request Type read-only in operational layouts; they are stamped by the application.

3. Check permission sets and actual user access

Review REDA AI Admin for administrators, REDA AI User for internal users and REDA AI Community User for the applicable community users. Assigned packaged sets can receive updated permissions; an upgrade does not create missing assignments or update your cloned/custom sets. For custom sets, review object access and field access for the table’s objects and fields. Required relationship fields follow the object’s access rules and may not have a separate editable field-permission entry.

MCP users need REDA AI User plus REDA AI MCP User; the latter also grants edit access to the two MCP configuration fields. Review that grant against your roles. The REDA AI MCP Guest set is for the endpoint’s Site guest user, not a substitute for authenticated users’ business-data permissions.

4. Check picklists and stale help text

On Agent Definition → Processing Tier, verify the API value Economy and its availability for applicable record types. On an existing Async Request → Request Type, verify Continuation and Bulk Turn. Add missing values using the exact API spelling. New fields bring their packaged values; existing subscriber configurations still need checking.

Older field help text can survive an upgrade. Review Processing Tier, Deep Reasoning, Conversation Window, Economy Tier Consent, prompt Context Objects, agent Default Context Objects, Agent ID and Persona. Use the current guides for behavior. In particular, Economy requests without consent are refused; replace any help text saying they automatically run as Efficient. Suggested consent help: “Confirms your organisation accepted the Economy processing disclosures. Without consent, Economy requests are refused. Select only after approval.”

5. Configure the features you intend to use

• Conversation continuity: set Conversation Window (Hours) explicitly where needed. Blank now uses the organisation default, initially 30 minutes; 0 prevents automatic resume. A positive value sets the window. Test record and no-record contexts; this setting does not delete history.
• Economy: request a reissued licence from REDA if your key predates Economy credentials, activate it through the normal licence process, and create/verify your approved active Remote Site Setting for https://api.deepseek.com. Review the processing terms before checking Economy Tier Consent on each intended agent. Missing consent or missing Economy credentials causes refusal. A provider-availability fallback is a different case and may be billed as Efficient. Ordinary Economy usage now uses 0.40× the base credit rate; file pre-extraction is full rate. Deep Reasoning is unavailable on Economy.
• Exclusive Prompts: add Exclusive Prompt and Summary to Custom Prompt. Create the shared prompt with no direct Agent Definition, mark it exclusive, and write a useful summary. Create an Exclusive Prompt assignment for each chosen agent, select the prompt, set Active and choose Always, On Context or On Demand (API values Always, On_Context, On_Demand). On Demand needs get_context_prompts available in the agent’s tool loadout. A prompt with a nonblank context scope no longer loads on a context-free turn: clear the scope only when that prompt is intended to apply everywhere. Check the result in Agent Runtime Preview, both with a record ID and with no record.
• MCP: configure the Site endpoint and Salesforce External Client App, approve the permission set in the app’s policy, assign users, then enable MCP only on the intended active Internal agents. Complete MCP Setup and the user guide. Installation alone does not connect an external assistant.

6. Merge tool and prompt configuration

In Setup → Custom Metadata Types → REDA AI Tool Definitions → Manage Records, compare existing subscriber-controlled values with the current REDA definitions. Preserve client customisations while merging these changes:

• create_task and create_event: Description and Input Schema must advertise field_values; create_event must include Internal;API in Channels if you want it on API agents.
• send_email: merge the current Description and Input Schema, including recipient_id, To/Cc handling and the recipient-versus-related-record guidance.
• schedule_action: merge the current Description. On older baselines also compare get_record against the current reference. Verify the required tools are actually offered by the intended agent.

Do not replace the entire tool record blindly: descriptions, schemas and channels are subscriber-controlled, while the packaged Requires Confirmation setting is developer-controlled.

Review _common:language-directive-api, _common:language-directive-external and _common:core-instructions-api. Merge updated language and human-recipient instructions while preserving your business rules. Seed Default Data can overwrite all shared default prompts; it is not a safe one-prompt update. API/external language should follow the incoming message, rather than the service or guest user’s language.

7. Review automation before running it

For bulk Flow invocations, check whether downstream elements expect an immediate answer: deferred processing returns an asyncRequestId instead of immediate response and credit output. Review every API agent configured with Execution Context = System: the patches make automated actions honor that setting where some previously ran in User mode. That can change effective object and field permissions; confirm it is intended before running the workflow.

For time-based actions, verify the scheduled processor is runnable and review its owner and existing pending work. New Pending or rescheduled actions are rejected when no runnable processor exists. Do not activate it merely to clear an error without reviewing what will run. Delayed jobs use the scheduler owner’s execution identity. Email-reply tracking and routine history/status updates are separate cases. Verify explicit datetime offsets in a disposable test.

8. Verify and hand over

Use an ordinary authorised user, with an administrator as a comparison. Confirm field visibility, allowed and denied record access, a harmless action proposal/decline/approval and its actual Salesforce result, conversation resume, and the features you enabled. Inspect a new credit transaction’s Processing Tier and an MCP message’s Via MCP flag. Older credit rows are not backfilled; do not interpret a historical blank as measured usage of a tier. Record the installed version, layouts/permissions changed, features enabled, test results and any remaining client-owned setup before wider use.

Related instructions

• Installing REDA AI
• Admin Overview
• Platform Settings Reference

Fields and access to review

The table covers fields introduced since the previous main release, including the intervening patches. Put administrator settings on the relevant configuration layout; keep system-stamped fields read-only in operational views. Use translated display labels in your org; API names below identify the field unambiguously.

Verify before extending access

Test as an ordinary user, with an administrator as a control: confirm approved and denied record access, conversation resume, a harmless proposal/decline/approval, and the actual Salesforce result. If adopting MCP, test sign-in, agent visibility, catalogue refresh and approved origins. Review an Economy transaction separately from file extraction and fallback. For schedules, verify the time and processor readiness before a disposable test. Check customised reports for Processing Tier. Escalate unexpected results before broader rollout. The release does not imply that your organisation's integration, permissions or data-policy review has already been completed.