To let an assistant use these tools, grant AI agents on the consent screen, or just
Agent compliance under its Fine-tune. The scope is agent_compliance.
Read tools work with Read only or Read & write access; write tools only with
Read & write.
Your assistant doesn't see these tools straight away. It finds them with search_tools or
describe_tools, and runs one with use_tool, passing its name and
arguments. Most assistants do this for you: just ask for what you want.
What your assistant is told about this group
The compliance controls around your agents. Policies: versioned sets of guardrails and escalation rules. Conduct rules: versioned rules written into the agent's instructions, with example replies, ready-made templates to copy, and each rule's check results. The rulebook: which agents carry which rules. The decision feed: each time the compliance gate checked something an agent was about to do, with the policies behind the decision, and each conversation's full audit trail. The review queue: actions the gate held for a person to approve (decide_tool_approval lets one go ahead: confirm with the user first). And the tools agents can use, with a record of every time they were used. get_sale_milestones scores a sale against its sales flow: which expected milestones were recorded, owed, missed or done out of order, by application or by any chat that worked it; waive_sale_milestone records an accepted gap (audit only, write-once), attest_sale_milestone and revoke_sale_milestone_attestation record (or withdraw) a manager's attestation of one of the firm's own milestones, and get_milestone_completion_report says how the org's sales ended against their sales flows, and get_disclosure_integrity_report how well the rating answers behind them are evidenced.
The descriptions below are the ones your assistant reads, so they're written for it. When a connection covers more than one
organisation, the assistant passes org (the organisation's slug, from list_orgs) to say which one.
get_compliance_overview
Read Compliance overview
Counts across the compliance estate: registry rules, policies, decisions, tool executions.
Arguments
| Field | Type | Description |
|---|---|---|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_agent_policies
Read List agent policies
Compliance policies (guardrail/escalation bundles) agents attach.
Arguments
| Field | Type | Description |
|---|---|---|
| archived | boolean | true lists the archived ones instead |
| limit | integer | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_agent_policy
Read Get an agent policy
One policy with its current version's rule sets and the agent versions that use it. usage.live is what blocks archiving it, usage.history what keeps it from being deleted for good.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
create_agent_policy
Write Create an agent policy
Create a compliance policy from registry rules (with config overrides) plus org guardrails and escalation rules.
Arguments
| Field | Type | Description |
|---|---|---|
| name required | string | |
| description | string | |
| screening_mode | string |
One of: "auto", "always", "never"
|
| rule_keys | array of string | Registry rule keys |
| rule_configs | object | Per-rule-key config overrides |
| guardrail_ids | array of string | |
| escalation_rule_ids | array of string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_agent_policy
Write Update an agent policy
Save a new version of a policy with the given rule sets (pass the FULL set you want: it replaces the old one). Draft agent versions pick up the new version; promoted versions keep theirs until a new version of the agent is promoted.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| name | string | |
| description | string | |
| screening_mode | string |
One of: "auto", "always", "never"
|
| rule_keys | array of string | Registry rule keys |
| rule_configs | object | Per-rule-key config overrides |
| guardrail_ids | array of string | |
| escalation_rule_ids | array of string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
archive_agent_policy
Write Archive agent policy
Archive the agent policy: it leaves lists and pickers and can't be newly chosen, while everything that already used it keeps working. Refused while anything live still uses it (the error says what, so switch those first). Undo with restore_agent_policy.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
restore_agent_policy
Write Restore agent policy
Restore an archived agent policy to lists and pickers.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
delete_agent_policy
Write Delete agent policy
Delete an ARCHIVED agent policy for good. Only possible when nothing, live or historical, has ever used it; otherwise it stays archived. Irreversible: confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
apply_agent_policy_pack_update
Write Update a policy to its pack's latest version
For a policy installed from a compliance pack with a newer released revision: replaces the policy's standard rules with the pack's, as a new policy version (agent versions still being set up move onto it; live agents take it when a new version of them is promoted). The org's own guardrails and escalation rules stay. Confirm first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
apply_rules_to_live_agents
Write Apply compliance rules to live agents now
A live agent runs the compliance rules it was promoted with, so a change to the firm's rule settings normally waits for the next promote. This applies the rules as they stand now to every live agent at once, recorded with the reason. Policy edits still need a new agent version promoted. Confirm first.
Arguments
| Field | Type | Description |
|---|---|---|
| note | string | Why the change can't wait for the next promote |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
compare_agent_policy_versions
Read Compare two versions of a policy
What changed between two versions of a compliance policy: its AI screening mode, rules added or removed, rules switched on or off, and settings changed (from and to). By version number: to defaults to the current version, from to the one before it. Lists the agent versions running each side.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| from | integer | |
| to | integer | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_conduct_rules
Read List conduct rules
Prompt-level conduct rules plus the platform template registry (topic, revision, which agents each is recommended for) with per-template usage.
Arguments
| Field | Type | Description |
|---|---|---|
| archived | boolean | true lists the archived ones instead |
| limit | integer | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_conduct_rule
Read Get a conduct rule
One rule with its current prompt text and examples, version history, whether its usage.live is what blocks archiving it, usage.history what keeps it from being deleted for good.template has a newer revision, and how each agent carrying it did on its checks.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
create_conduct_rule
Write Create a conduct rule
Create a custom conduct rule (goes into agents' system prompts), optionally with example replies that keep and break it. To start from a platform template, use adopt_conduct_rule_template instead.
Arguments
| Field | Type | Description |
|---|---|---|
| name required | string | |
| description | string | |
| prompt_text required | string | |
| examples | array of object | Example replies, at most 6 |
| examples[].kind required | string |
One of: "good", "bad"
|
| examples[].text required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_conduct_rule
Write Update a conduct rule
Edit a conduct rule. Changing prompt_text or examples saves a new version, which draft agent versions pick up (promoted versions keep theirs). Examples only change together with prompt_text: to change only the examples, pass the current text too.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| name | string | |
| description | string | |
| prompt_text | string | |
| examples | array of object | Example replies, at most 6 |
| examples[].kind required | string |
One of: "good", "bad"
|
| examples[].text required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
archive_conduct_rule
Write Archive conduct rule
Archive the conduct rule: it leaves lists and pickers and can't be newly chosen, while everything that already used it keeps working. Refused while anything live still uses it (the error says what, so switch those first). Undo with restore_conduct_rule.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
restore_conduct_rule
Write Restore conduct rule
Restore an archived conduct rule to lists and pickers.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
delete_conduct_rule
Write Delete conduct rule
Delete an ARCHIVED conduct rule for good. Only possible when nothing, live or historical, has ever used it; otherwise it stays archived. Irreversible: confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
adopt_conduct_rule_template
Write Copy a conduct rule template
Copy a platform conduct rule template (see list_conduct_rules) into an editable, versioned rule of the org's own, with its example replies. The copy remembers the template revision, so a later revision can be applied with update_conduct_rule_from_template.
Arguments
| Field | Type | Description |
|---|---|---|
| template_key required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_conduct_rule_from_template
Write Update a conduct rule to its template's latest revision
For a rule copied from a platform template whose template has a newer revision (template_update_available): replaces the rule's wording and examples with the template's, as a new version. Overwrites any edits made to the copy: confirm first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
run_conduct_rule_checks
Write Run a conduct rule's checks
Starts a compliance check run of the rule's written checks against every agent that carries the rule (its live version, else its draft), on web chat. Uses AI credit: confirm first. Returns the runs started; follow them with get_check_run.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_rulebook
Read Get the rulebook
Every agent against every rule it could carry (the org's conduct rules, platform templates attached as is, and policies): whether its live or only its draft version carries each, and which recommended conduct rule templates its live version is missing (add them in a new agent version with create_agent_version).
Arguments
| Field | Type | Description |
|---|---|---|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_policy_decisions
Read List policy decisions
The compliance gate feed, as the dashboard shows it: every screening, send and tool decision (each tool call once, with how it ended) and the policies in force for each, with per-decision counts. decision: require_approval surfaces the human-review queue.
Arguments
| Field | Type | Description |
|---|---|---|
| action_group | string |
One of: "screen_inbound", "screen_outbound", "messaging", "tool"
|
| decision | string |
One of: "allow", "allow_with_obligations", "deny", "require_approval"
|
| agent_version_id | string | |
| rule_key | string | |
| limit | integer | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_policy_decision
Read Get a policy decision
One gate decision: the triggered rules, the policies in force (pinned to its agent version, verified against the fingerprint recorded with it) and where each rule came from, every rule in force with its settings at the time and the changes it made, the build it ran on, its place in the org's hash chain (verified, still being secured or broken), how long it's kept, the AI grader's calls behind a content check, any tool-approval request, and for a tool call every step (arguments check, permission, run, result check).
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_conversation_audit
Read Get a conversation's audit
The full audit trail of a conversation and every other chat with the same customer, in order (for a conversation with no customer, the chats it was handed along): channels and agents per chat, the sales scripts the sale ran on (each with its sales flow and product release), the controls in force (agent versions, conduct rules, policies, tools, platform protections), consent and vulnerability, and every message, check, tool call, escalation, approval and human action with the reasons recorded at the time. Includes a fingerprint of every record it's built from.
Arguments
| Field | Type | Description |
|---|---|---|
| conversation_id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_sale_milestones
Read Get a sale's milestones
Where a sale stands against its sales flow: every milestone the firm expected of it (marketing preference asked, health data consent, right to cancel sent, key documents sent, needs summary confirmed, quote accepted, identity captured, Direct Debit details captured, first premium collected, policy bound, handed off), each achieved, owed, waived or revoked, with when it happened, on which channel and the ledger row it rests on. Says which required milestones are still owed, which were missed once the sale ended, which happened out of order and which rest on evidence reconstructed by a backfill rather than recorded at the time. Give application_id, or conversation_id for any chat that worked the sale. The same scorecard the conversation's audit tab shows; read-only.
Arguments
| Field | Type | Description |
|---|---|---|
| application_id | string | The policy application (the sale) |
| conversation_id | string | A chat that worked the sale, instead of application_id |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
waive_sale_milestone
Write Waive a sale's milestone
Record that the firm accepts a sale didn't reach a required milestone, and why (get_sale_milestones lists them). The scorecard, summary and audit then show it waived, by whom and why. For the record only: a waiver never lets the sale proceed past a step (the gate and the agent still hold it to the milestone). It can't be undone: the milestones ledger is write-once. Refused for an outcome, a milestone achieved, not in play, not required or already waived. Confirm the reason with the operator first. Managers only.
Arguments
| Field | Type | Description |
|---|---|---|
| application_id required | string | The policy application (the sale) |
| milestone required | string | Catalogue milestone key, e.g. key_documents_sent |
| reason required | string | Why the gap is accepted (kept with the waiver) |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
attest_sale_milestone
Write Attest a firm milestone on a sale
Record that a sale reached one of the firm's own milestones (the firm requirements on its sales flow's components, keys firm_...: things the platform can't observe, such as a vulnerability check), with a note of what was done. get_sale_milestones lists the sale's milestones; the firm's own have evidence "attested". The scorecard, summary and audit then show it achieved, by whom and why. Firm milestones never hold a sale back. revoke_sale_milestone_attestation withdraws it. Refused for a platform milestone, one not in the sale's sales flow or not in play, one already attested, and a compliance check's sale. Confirm the note with the operator first. Managers only.
Arguments
| Field | Type | Description |
|---|---|---|
| application_id required | string | The policy application (the sale) |
| milestone required | string | The firm milestone's key, e.g. firm_vulnerability_check |
| note required | string | What was done (kept with the attestation, at most 1000 characters) |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
revoke_sale_milestone_attestation
Write Revoke a firm milestone's attestation
Withdraw the attestation of one of the firm's own milestones on a sale (attest_sale_milestone), and say why. The milestones ledger is write-once, so the attestation stays on record and the scorecard shows the milestone revoked; it can be attested again. Refused unless the milestone is attested on the sale now, and for a compliance check's sale. Confirm the reason with the operator first. Managers only.
Arguments
| Field | Type | Description |
|---|---|---|
| application_id required | string | The policy application (the sale) |
| milestone required | string | The firm milestone's key, e.g. firm_vulnerability_check |
| reason required | string | Why the attestation is withdrawn (at most 1000 characters) |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_milestone_completion_report
Read Milestone completion report
How the org's sales ended against their sales flows: the sales that ended (policy bound, handed to the insurer or to a person) in the period, how many met every required milestone, how many milestones were waived and the one missed most, broken down by product, channel and week, and the sales that ended with gaps (most recent first, with links to the application and its audit). The period is at most 24 months (capped says when it was shortened) and by_week the most recent 104 weeks. Complete counts sales that met every required milestone with nothing waived; complete_with_waivers counts the rest that met them only with waivers. The compliance overview page shows the same.
Arguments
| Field | Type | Description |
|---|---|---|
| from | string | First day, YYYY-MM-DD (24 months before `to` when omitted) |
| to | string | Last day, YYYY-MM-DD (open when omitted) |
| product_id | string | Narrow to one product |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_disclosure_integrity_report
Read Disclosure integrity report
How well the rating answers behind the org's sales are evidenced, for a reinsurer or an auditor: of the sales that ended in the period, how many met "rating answers evidenced" (every answer the price rests on given by the customer, on a ledger record that still matches it), and across them how many answers the agent recorded without the customer's own words (unverified), an assistant app sent (host_reported: the customer's words aren't seen), were never recorded (unrecorded) or no longer match their record (mismatched: changed outside the platform), and how many changed after the customer was given a price. Each sale is listed (most recent first, up to 5,000) with links to the application and its audit, and ledger says whether the milestones ledger's hash chain is intact. Counts are as of each sale's last milestone stamp; a sale's audit reads its answers live. The period is at most 24 months (capped says when it was shortened).
Arguments
| Field | Type | Description |
|---|---|---|
| from | string | First day, YYYY-MM-DD (24 months before `to` when omitted) |
| to | string | Last day, YYYY-MM-DD (open when omitted) |
| product_id | string | Narrow to one product |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
decide_tool_approval
Write Approve/deny a gated tool call
Decide the human-review request behind a require_approval decision and resume the conversation. Approving lets the gated action proceed: confirm with the operator first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | Policy decision id |
| status required | string |
One of: "approved", "denied"
|
| note | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_agent_tools
Read List agent tools
The org's agent-callable tools (platform and custom) with confirmation/approval flags.
Arguments
| Field | Type | Description |
|---|---|---|
| limit | integer | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_agent_tool
Read Get an agent tool
One tool with its parameter schema and the agent versions using it.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_tool_executions
Read List tool executions
Agent tool-call history, newest first. Filter by status, tool name or agent.
Arguments
| Field | Type | Description |
|---|---|---|
| status | string |
One of: "pending", "success", "error"
|
| tool_name | string | |
| agent_id | string | |
| limit | integer | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_tool_execution
Read Get a tool execution
One execution in full: arguments, result payload and error.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_policy_rules
Read List policy rules
Every rule the org's policies can use (content, escalation, and delivery and action rules), with the org's own default overrides and how many of its policies use each. Retired rules, and the software-sales escalations the org doesn't use, are left out.
Arguments
| Field | Type | Description |
|---|---|---|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_policy_rule
Read Get a policy rule
One policy rule: its defaults, the org's overrides, which agent policies use it and its ten most recent triggers.
Arguments
| Field | Type | Description |
|---|---|---|
| key required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_org_policy_rule
Write Set the org's defaults for a policy rule
Sets the org-wide defaults for one rule, used by every agent policy that doesn't override them: config keys come from the rule's default_config (blank or default values fall back), and a key listed in its config_choices takes one of those values only; enabled false switches the rule off org-wide. Agent versions still being set up use the change at once; a live agent takes it when a new version of it is promoted, or when the rules are applied to live agents now (apply_rules_to_live_agents). Confirm first. Manager or owner.
Arguments
| Field | Type | Description |
|---|---|---|
| key required | string | |
| config | object | |
| enabled | boolean | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |