Agents you can trust with your Salesforce answers.
For Agentforce agents and Claude. We clean and define the data, set the access, test the answers, and keep them right.
Start with a readiness check
We review your data, access and the questions your team asks of Salesforce, then tell you what to fix first and where each job should run.
Start with a readiness check →Agents built in your org
14 agents and 6 packs for every team, plus Claude for your people.
Browse the agent catalogue →Three places to start.
Pick the one that matches the question you are being asked. Each has a fixed scope and a price in USD.
Agentforce Readiness Assessment
Find out where AI will pay off, which tool should do each job and what has to be fixed first. Includes a baseline answer test.
$2,500 / $6,500 / $9,500 USDSee the assessment →Data Cleanup and Dedupe
Dedupe and merge the records agents will read, fix the fields they depend on and set rules so it stays clean.
$3,500+ USDSee Data Cleanup →Trusted Answers
Claude on your Salesforce data, with answers you can check. We handle data fixes, access and write policy, skill tuning and an answer test set.
$4,500 / $9,000 / $13,500+ USDSee Trusted Answers →Answers you can trust, by fixing what's underneath
An agent or Claude is only as good as the data, access and rules under it. We work through five steps, and each one has a check you can see.
- 01
Data ready
Dedupe and merge existing records, fix the fields agents read, set duplicate and validation rules so it stays clean. Where an agent answers from knowledge, PDFs, SharePoint, Drive or transcripts: knowledge grounding through Data 360 (Data Library and a retriever) inside your free Data 360 allowance where it fits.
How we checkDuplicate rate and field fill rate against agreed thresholds (for example under 5% duplicates, 85%+ filled), measured before and after. For grounding: an answer test set scored for correct source and correct answer.
- 02
Access ready
Agent users, permission sets, sharing review, field-level security. For Claude: Salesforce MCP scopes and write policy, tested as real users.
How we checkA test run as each agent user and two real user profiles shows exactly what each can see and change.
- 03
Cost guardrails
A Flex Credit run-rate forecast per agent, one identity per agent and Digital Wallet alerts, plus a simple on/off switch per agent that your admin controls. If you already have MuleSoft Agent Fabric (a separate MuleSoft product), we set up and tune its budgets, dashboard and kill switch instead of building our own.
How we checkA monthly cost forecast per agent, and a tested kill switch that stops an agent in under a minute.
- 04
Tested
A scenario suite per agent, written from your team's real questions and confirmed true answers, with pass thresholds agreed with you. Run in Agentforce Testing Center (or Agent Evals in MuleSoft Agent Fabric once available, if you have it), with French scenarios where needed. Re-run after each Salesforce release.
How we checkA pass rate against the agreed scenarios before go-live and after every release.
- 05
Governed and monitored
Named owners, a change process, an agent register, a monitoring dashboard (sessions, errors, credits, outcomes).
How we checkYour admin can pause, adjust and report on every agent without us.
Already piloting? Make it production-ready.
Four fixed-price services for agents and Claude connections you already have. Each one stands alone.
Knowledge Grounding with Data 360
For agents that answer from your content. Agentforce needs Data 360 for that, even for Knowledge articles. We check whether your free Data 360 allowance covers it, ground one source, and test the answers before your customers see them.
$6,500 USDSee the details →Access and Permission Review
Agent users, permission sets, sharing and field-level security reviewed before any agent or Claude connection. Claude sees whatever the user can see.
$3,000 USDSee the details →Agentforce Cost Guardrails
For agents you already have: a Flex Credit run-rate forecast per agent, one identity per agent, alerts and a simple on/off switch your admin controls.
$3,500 USDSee the details →Agent Test Suite
For agents you already have: scenarios and answer tests written from your team's real questions, pass thresholds agreed with you, and a re-run after every Salesforce release.
$2,500 USD per agentSee the details →Agents built in your org, for every team.
Three of our fourteen agents. Each runs inside your Salesforce org, follows your fields and rules, and is tested against scenarios we agree with you before launch.
Data and platform
Find and fix duplicate and incomplete records in any Salesforce object, with changes approved by your team.
Built to order in your org. Not a template.Data and platform
Know exactly what's risky, unused and fragile in your Salesforce org, get a prioritized list of what to fix first, and give your team someone to ask "how does this work here?"
Built to order in your org. Not a template.Service
Every new case classified, checked for duplicates and sent to the right team in minutes, with the reason written on the case.
Built to order in your org. Not a template.Not in the catalogue? We build custom agents on Agentforce or other models, scoped after the assessment.
Agents for your records. Claude for your people.
The dividing line is people or records. When someone wants help with their own deals and accounts, Salesforce in Claude (Salesforce's own plugin for Claude) is often the right tool. When a job has to happen on every record without anyone asking, it needs an agent inside Salesforce.
Three of these four you can buy without us. We check what you already own first, then recommend whichever fits.
Native Agentforce
A record event, schedule, user or customer
Fit report, configuration on request
Salesforce in Claude
A person asking, or a task they schedule
Trusted Answers: access review, write policy, skill tuning, teams beyond sales, training
AgentExchange
Same as native
Fit report, mapping and testing
ArkeAgents custom agent
A record event or schedule, on every record
We build it: the 14 agents in our catalogue
If Salesforce already does it, we'll say so.
Not sure where AI fits? Start with the Agentforce Readiness Assessment.
We review your org, score the questions your team really asks of Salesforce, shortlist the jobs worth doing and recommend where each one should run, including the options we don't sell. You own the report either way.
We never change your org during the assessment.
See the full assessment →Baseline answer test
We take a short set of questions your team really asks of Salesforce and score the answers at the start. Every engagement scores the same questions at the end, so you see the before and after.
The fee is credited against any ArkeAgents engagement signed within 60 days, up to $2,500 USD and never more than half the engagement price.
Book the assessment →How we work.
Built faster by our engineers with AI assistance. Every change is reviewed and tested before it reaches production.
- STEP 01Assess
We start read-only. We review your data, automation and security, and the jobs your teams repeat, then shortlist where AI would help.
- STEP 02Fit
For each job we decide with you where it should run: Native Agentforce, Salesforce in Claude, an AgentExchange listing or a custom agent. Each decision and its reasons go in your fit report.
- STEP 03Build
Our engineers build in your sandbox, bound to your fields, rules and approval steps. Every agent is tested against scenarios we agree with you before it reaches production.
- STEP 04Launch
Agents start in report-only or shadow mode on your real data. Your team sees what each one would do before it does anything.
- STEP 05Tune
Once an agent is live, we adjust its rules and thresholds against the targets in your scope. Your admin holds the switch to turn it off.
The team that assesses is the team that builds.
The engineers on your first call are the engineers who review your org, write the fit report and build in your sandbox. You are never handed to a delivery team you haven’t met.
Scope and price are agreed before we start, never after. Your AI and data-use terms are confirmed before any tool connects to your org.
Arke was the swift messenger of Greek myth: a messenger who acts for you, inside your rules. Why ArkeAgents →
Thirty minutes tells us
if we're a fit.
Bring one process or one piece of Salesforce work. We'll tell you where it should run and what we would build. Written scope and fixed price within 48 hours.