Every Kobal agent, accounted for
Kobal’s agents work out of sight, behind your feed and your Slack. Command Center is the one place you see them all: what each one ran, what it cost, how good its work is and anything that failed.
- Agents
- 15
- Running now
- 2
- Runs
- 918
- Cost
- $4.12
- Quality
- 4.4
- Needs attention
- 2
Finding agents · 13
- Alert / AnomalyEvery 15 min09:304.6
- Risk RadarHourly09:004.2
- PrioritisationWhen findings change09:214.5
- Root CauseIn investigationsSat 10:404.1
- Business ImpactWith other findings09:004.7
- ClusteringNightly02:004.3
- Taxonomy & PlaybookNightly02:10Not enough history yet
- Report / DigestWeekly08:004.8
- Strategic SignalsMonthly1 Sep4.4
- Ask-DataWhen asked09:124.5
- Decision AssistantWhen askedThu 16:054.0
- QA & CoachingDormantWould score conversations against your rubric. Waiting for a rubric and who handled each ticket.
- Demand & CapacityDormantWould forecast next week’s tickets against your team’s hours. Waiting for team schedules and SLA targets.
Engine · writes and sends for the others
- DraftingWhen another agent asks09:02—
- Workflow ExecutionOn approval09:21—
Running now
- Alert / AnomalyChecking the billing queue
- Risk RadarRescoring 9 accounts
Needs attention2
- Flag Acme Corp at riskCRM said no · Workflow Execution
- 2 waiting for your yes
You never have to learn their names
Every alert, draft and report Kobal sends carries a small label naming the agent that made it. Your team can ignore it, and when you want to know more, it opens that agent in Command Center.
- Plan
- Pro
- Renews in
- 30 days
- Tickets
- 3
- Started
- After Friday’s release
Payment went through on Friday night and we still can’t get into the workspace.Review evidenceAsk Kobal
Finds accounts at risk and says why.
- Watches
- All plans · billing, cancellations, competitor mentions
- Checks
- Hourly · last run 09:00
- On its own
- Nothing yet. Every action waits for your yes.
Fifteen agents, one job each
Kobal ships a fixed set of agents. You switch each one on, choose what it watches and decide what it may do on its own.
Alert / Anomaly
Warns you when today’s tickets don’t look like a normal day.
Risk Radar
Finds accounts at risk and says why, before they cancel rather than after.
Prioritisation
Puts what needs you first by business impact, not by ticket count.
Root Cause
Explains why something changed by lining the tickets up against your releases.
Business Impact
Puts a revenue figure on each problem, from your CRM and billing.
Clustering
Turns broad tags into precise problems: “login” becomes “reset emails not reaching Outlook users”.
Taxonomy & Playbook
Keeps your categories and macros clean, and spots questions your team answers differently.
Report / Digest
Writes your weekly report and sends it to Slack or email, every figure linked to its tickets.
Strategic Signals
Pulls out what matters for product: feature requests, competitor mentions and pricing friction.
Ask-Data
Answers your questions about your support data, with the tickets behind each answer.
Decision Assistant
Weighs your options when you ask, and recommends one with the trade-offs laid out.
QA & Coaching
Scores conversations against your rubric and turns the gaps into coaching.
Demand & Capacity
Forecasts incoming tickets against your team’s hours, before your SLAs slip.
Drafting
Writes the message, reply or report another agent needs, ready for your edits.
Workflow Execution
Carries out what you approved in your tools, and records what happened.
Know what every agent costs
Every run is recorded with what it cost, and each agent’s share sits against your total, so one running away is plain to see. Checks cost next to nothing and the costly writing only runs when a rule fires, so a busy weekend barely moves the number.
- Every ticketTagged once, by a small model.
- Every rule crossedWritten up by a larger model. Repeats join the same write-up.
- Your own toolsTheir calls land here too, with their own cost.
How investigations stay within their limitsHow your own tools call Kobal
- Tickets
- 2123.5× a normal weekend
- Cost
- $1.621.2× a normal weekend
- Root Cause1 investigation, 12 steps$0.46of its $0.50 limit
- Clustering3 nightly runs$0.38
- Tagging212 new tickets, small model$0.21
- Alert / Anomaly250 checks, no model · 2 write-ups, larger model$0.09
- Risk Radar62 checks · 1 write-up$0.06
- Report / Digest1 weekly report$0.05
- Your own tools6 API calls$0.02
- 8 more agents$0.35
Trust it has to earn
Each agent is scored on how often you approved, edited or dismissed its work. When one kind of action keeps earning your yes, Kobal shows you the record and asks whether it can run on its own.
How approvals and rules workQuality
- Five weeks agoNot enough history yet6 reviews
- Two weeks ago4.324 reviews
- Monday 09:204.641 reviews
Kinds of action
- Warn in SlackRuns on its own
- Hand off to a team in SlackWaits for your yesKobal asksApproved 38 of 41 · edited 2 · dismissed 1. Let hand-offs run on their own?
- Open an investigationWaits for your yes
Two agents never get a score. The ones that write and send for the others have no work of their own to judge, so theirs counts toward the agent that asked.
What they doBuilt to be checked
It never promotes itself. It can ask for more room; only you can give it. It won’t grade itself early. No score appears until your reviews can back one. It doesn’t fail in silence. A failed run goes to the top of Needs attention, with what came back. It can’t outspend you. Every agent works within a budget per run and a monthly cap you set. It says what it can’t do yet. A greyed agent tells you what it would do and what it’s waiting for.
Questions, answered
Can't find what you're looking for?
Visit our Help Center.
Stay in charge of every agent
We’ll walk you through what each agent watches, what it may do on its own and what always waits for you.
Get early accessWorks on top of the helpdesk you already use. No migration.