For AI agents: a documentation index is available at https://docs.coverbase.com/llms.txt. This page is also available in markdown by appending .md to the URL.
This guide is part of the User Guides collection. It covers the Risk entry in the left navigation: the register, the chain, scenarios, exposure and the board report. The Geography tab has its own guide, Risk geography map. Vendor-level risk scores and how services roll up into them are in the Admin and setup guide.
Risk is an optional module. If you do not see Risk in the left navigation, ask your Coverbase representative to turn it on.
Where it lives
Click Risk in the left navigation. Six tabs run across the top, in the order a risk officer reads them:
Two buttons sit to the right of the tabs. Ask the register opens a question box that answers from the register, the chain and the signals, and cites the records it used. Configuration opens Configuration → Risk, covered below.
Reading the register
The Register tab. Summary tiles across the top, the table with its filters, and the side panel with domains, sources and the last agent run.
Appetite is read from the domain’s Max Residual (default 12). A residual above it is Outside, a residual within two points of it is Near, and anything lower is Inside. A risk without a domain uses the default and rolls up as “No domain”.
The filter bar narrows by domain, owner, institution, signal source, status scope (Active risks, Closed risks, All risks) and free text, and sorts by residual, inherent, review date or last updated. Matrix switches the table to the 5x5 heat map, with inherent and residual dots and the tolerance line drawn on it. Every filter is in the URL, so a filtered view is a link you can send.
The side panel lists By Risk Domain (count and how many are outside appetite) and Where Risks Come From (signals per source module). Clicking a row filters the table. Underneath, the panel shows when the agent last ran and what it did, with a Run agent now button.
Statuses
Archiving is separate from closing. Archive removes the risk from the register and its scores. Its signals stay on their source records.
Creating a risk
1
Open the form
Click New risk in the filter bar. From a vendor, service, contract or finding page, the Risks card has Add risk, which opens the same form with that record already in scope.
2
Write the statement
Statement is what a board reader sees. The placeholder asks the right question: what could hurt the organization, and how?
3
Classify and assign
Pick a Risk Domain (its appetite decides the tolerance), a Category, an Owner (a person or a department) and, if your org has institutions, an Institution or Group.
4
Score it
Pick Likelihood and Impact from your org’s five levels; the inherent score updates as you choose. Then pick Residual Likelihood and Residual Impact for the position after treatment. Leave them at Unscored only if you have not treated the risk yet.
5
Treatment and scope
Choose a Treatment (No treatment, Mitigate, Accept, Transfer, Avoid), a Review Date, and write the Treatment Plan: what is being done, by whom, and the residual target. Add the vendors in scope. Click Create risk.
Working a risk record
Click a row to open the record. The URL becomes/risk/register/<id>, so it is a link too.
A risk record. Scoring against the domain tolerance, the owner and treatment actions, the chain position, the agent's rationale, the signals and the history.
In the Chain shows how many business units and customer channels sit behind the risk’s vendor, with Open in Chain. A risk not tied to a chain node says so.
Agent rationale appears on agent-created risks and on any risk the agent re-scored. Explain more expands it. If the agent applied a re-score, Undo restores the previous levels and journals the reversal.
Signals on a risk
Signals lists every attached record with its source icon, its severity, and its status. A status that changed since attachment reads “from → to”. A signal that closed says Closed, a dismissed one Dismissed, and one whose source was archived Source removed. Nothing is ever deleted from this list, so a finding that was open when the risk was raised still shows on the timeline after it resolves.- Attach signal opens evidence the agent found in other modules but did not attach. Attaching keeps a reference, never a copy. Suggest more asks the agent to look again.
- The x on a signal dismisses it. The source record is untouched.
Signal sources and thresholds
A signal is attached only when its source is enabled under Configuration → Risk → Signal Sources and the record clears the threshold set there. Every source is on by default except Assessments.
Turning a source off stops new attachments. It never removes signals already on a risk. Signal statuses refresh when the source module changes and again in a nightly pass.
The agent
The agent runs every night, about five minutes after a qualifying change in another module, and whenever someone clicks Run agent now (on the register side panel or under Configuration → Risk → Agent). One pass refreshes signal statuses, scans the sources, attaches new evidence to the open risk on the same vendor and category (or proposes a new risk), suggests re-scores, and flags stale risks. Two settings under Agent decide how much it does on its own:- New risks: Propose for review (default) files each new risk as Proposed for a person to accept. Create directly opens them as Open risks.
- Re-scoring: Suggest files a re-score suggestion. Apply and notify writes the new levels and journals the change; you can still Undo on the record.
Deciding on what the agent proposes
Click the Proposed by agent tile. The inbox has four tabs.
The agent inbox. Each card carries the reasoning, the suggested scoring and owner, and the evidence, with the decision buttons underneath.
Dismiss asks why: Not a risk for us, Duplicate, Wrong scoring, Already treated or Other. The reason is fed back to the agent and recorded in history.
Each card shows the agent’s confidence, its rationale under Why, and the signals under Evidence. A proposal the agent files with no signals of its own still cites the records it read.
Chain
The Chain tab following Data. Business units on the left, then services, vendors and fourth parties. Red nodes are single points of failure.
Narrow with Data scope, Business unit, Depth and Show only critical paths. A column with too many nodes collapses the rest behind +N more; Show all expands it. In the legend, a red edge has no contractual protection, a ring marks a node with an open register risk, and a red node is a single point of failure.
Click a node to see what it depends on, what it serves, its contract (liability cap, SLA, DPA in place or not), and what is open against it (findings, obligations). Simulate outage runs an event scenario on it. Single points lists every single point of failure by reach, with how many alternatives are on record.
Simulate
The Simulate tab. Build an event, change or offboarding scenario on the left, or type the question under Or ask. The result fills the panel on the right.
- Event: an Outage, Breach, Insolvency, Sanction or Market exit on a Subject from the chain, for a Duration, starting in Business hours, weekday, Overnight or Weekend.
- Change: Describe the change in a sentence and name the Vendors involved and Business units involved. The result lists what it touches: control sets, register risks, obligations and open questions for stakeholders.
- Offboard: a Vendor to offboard. The result is a brief: termination terms on file, dependencies, open findings and obligations, and alternatives.
The engine never fills in a number you have not recorded. If an estimate needs an input the org does not hold (a liability cap, a business unit’s revenue), the loss reads Estimate withheld with the reason, and Improve it lists what to record. Confidence is derived from how many inputs are present, never asserted.
Exposure
The Exposure tab. One gauge per domain, the loss curve with its inputs coverage, the cushion table and the concentration bars.
- Appetite by domain: one gauge per domain plus “No domain”, with the count outside appetite, the appetite statement, the average residual and a 90-day trend (rising, flat, falling). Click N outside to open those risks in the register.
- Probable annual loss: the loss curve with the appetite line, the Method chip (change it under Configuration → Risk → Quantification), Inputs present per input type, the confidence, and Vendors without inputs. Only vendors with a residual score and a sized exposure are modeled; the rest are named, not defaulted.
- Cushion analysis: per contract, the ACV, liability cap, insurance (none on file or not extracted) and the unprotected amount. Extract from contract runs clause extraction on rows missing a cap. Unknown stays unknown.
- Concentration: shared fourth parties, business units on one vendor, and geography, each linking into the chain.
Report
The Report tab for a quarter. The four indicator tiles, movers and appetite breaches, the KRI table by domain, and the narrative.
Configuring the module
Open Configuration → Risk, or click Configuration on the Risk page. Each section saves on its own.
Signal Sources in configuration. One switch and threshold per source module.
How the other modules connect
Vendor, service, contract and finding pages each carry a Risks card listing the register entries scoped to that record, with their residual score and appetite state, Add risk, and View in register, which opens the register filtered to that record (“Scoped to this vendor”). Proposed risks show as Proposed by agent on the card. An intake request for a vendor you already hold shows the same card, as Open risks on this vendor, beside the intake questionnaire until the request is submitted. The reverse direction is the signal table above: findings and their commitments, Radar signals and alerts, contract components and clause reviews, obligations, sanctions screening, the financial health score, supplier sites, Inspect evaluations and assessments all reach the register as signals on the risks that cover them. Charts on your dashboards can slice the register by domain, status, appetite, owner and institution. Workflow automations have Create risk and Update risk actions (see Workflow templates). A risk that needs a person (a proposal, an overdue review, or an unowned risk outside appetite) also appears once in the work queue, assigned to its owner, or failing that the owning department’s lead or a domain reviewer.Who can do what
Risk has its own permission resource, Risk register, with these grants. Members hold them by default; a missing button means the grant is missing from your role.
Saving anything under Configuration → Risk needs the organization settings permission, which Admins hold. Roles are managed as described in Permissions and roles.
Notifications
Two notification types live under the Risk category in your personal notification settings. Both group one agent run’s output into a single email.
Each email carries a button that opens the risk. Delivery settings are covered in Email and notifications.
Work the register from MCP or the API
You can keep the register, run the agent, simulate and read the report from an MCP client or with an API key, as well as in the dashboard. An MCP client reads withget_report(report_name=...) (risk_register_query, risk_detail, risk_matrix, risk_chain, risk_exposure, risk_board_report). It writes with mutate(entity="risk", action=...), mutate(entity="risk_scenario", action=...), mutate(entity="risk_chain", action="rebuild") and mutate(entity="risk_narrative", action="save").
The API routes sit under /v1/risks, /v1/risk-scenarios and /v1/risk-chain, with GET /v1/risk-exposure and GET /v1/risk-report. See the Risk API.
Before an MCP client sends a write, it shows you the change and waits for you to confirm it. The MCP tool reference lists every field.
To have the AI write a report’s narrative, an MCP client calls mutate(entity="risk_narrative", action="start_job") and reads the result with get_report(report_name="text_job"). The API starts it with POST /v1/risk-narrative/jobs and returns it from GET /v1/text-jobs/{job_id}. See AI narratives and suggestions.
Archiving a risk needs risk:archive:org, which service accounts do not hold. Changing the module configuration with PATCH /v1/risk-config needs organization settings permission, so use a key created by a person whose role holds it.
Troubleshooting
Related
Risk geography map
The Geography tab: supplier sites, country counts and the vendor filters.
Admin and setup guide
Vendor risk scales, service roll-up and the rest of the organization setup.
Contract components
The DPA, SLA and liability cap slots that feed component-gap signals and the cushion table.
Email and notifications
Where to turn the two risk notifications on or off.