Agents

Every Agents tool, with the scope it needs, its arguments and the API operation it calls.

15 tools

List agent activity

Read only

atlas_agents_activity_list

List recent agent execution activity rows, newest first, one page at a time: queued / running / completed / failed / approval-needed / max-steps-exceeded. Filter by agentId or status; limit is 1 to 100. Returns { items, nextCursor }; pass nextCursor back as cursor for the next page.

Scope: workspace:read
Calls: GET /agent-governance/activity
Example prompt: What has the planner agent been doing this morning?

Arguments

Field	Type	Required	Description
`agentId`	`string`	No	Filter to a single agent id.
`status`	`"queued" \| "running" \| "completed" \| "failed" \| "approval-needed" \| "max-steps-exceeded"`	No	Filter to a single execution status.
`limit`	`integer`	No	How many rows to return, 1 to 100. Defaults to 50.
`cursor`	`string`	No	The nextCursor of the previous page. Leave it out for the first page.
`order`	`"asc" \| "desc"`	No	asc or desc. Newest first (desc) by default.

Create agent approvals

Changes data

atlas_agents_approvals_create

Create a pending approval (planner / governed template / replay drill source). Owner/admin only; idempotent on retry.

Scope: workspace:manage
Calls: POST /agent-governance/approvals
Example prompt: Open an approval for the planner to send the renewal email.

Arguments

Field	Type	Required	Description
`agentId`	`string`	Yes	Agent id requesting the approval.
`runId`	`string`	No	Optional originating run id (links the row back to a planner trace).
`source`	`"agent_planner" \| "governed_template" \| "system_replay_drill"`	Yes	Where the approval was raised from.
`actionKind`	`string`	Yes	Domain action being approved (e.g. send_email, mutate_crm_deal).
`summary`	`string`	Yes	Human-readable summary surfaced to the reviewer.
`payload`	`object`	No	Action payload the approval gates - replayed verbatim if approved.

List agent approvals

Read only

atlas_agents_approvals_list

List the approval queue, newest first, one page at a time. Owner/admin only. Filter by status (PENDING / APPROVED / REJECTED) and / or agentId. Returns { items, nextCursor }; pass nextCursor back as cursor for the next page.

Scope: workspace:read
Calls: GET /agent-governance/approvals
Example prompt: What pending agent approvals are waiting on me?

Arguments

Field	Type	Required	Description
`status`	`"PENDING" \| "APPROVED" \| "REJECTED"`	No	Filter to a single approval state.
`agentId`	`string`	No	Filter to a single agent id.
`limit`	`integer`	No	How many rows to return, 1 to 100. Defaults to 50.
`cursor`	`string`	No	The nextCursor of the previous page. Leave it out for the first page.
`order`	`"asc" \| "desc"`	No	asc or desc. Newest first (desc) by default.

Replay agent approvals

Changes data

atlas_agents_approvals_replay

Replay an APPROVED approval payload as a new agent run. Owner/admin only. Optional runId / reason ride along on the new replay record.

Scope: workspace:manage
Calls: POST /agent-governance/approvals/{id}/replay
Example prompt: Replay approval ap_42 as a drill.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Approved approval id to replay.
`runId`	`string`	No	Optional new run id to associate with the replay attempt.
`reason`	`string`	No	Optional replay rationale (recorded on the run row).

List agent approvals replayable

Read only

atlas_agents_approvals_replayable_list

List APPROVED approvals that are still safely replayable (within retention window, payload still resolvable). Powers the replay-drill picker.

Scope: workspace:read
Calls: GET /agent-governance/replayable-approvals
Example prompt: Show me the last 50 approvals I can replay.

Arguments

Field	Type	Required	Description
`limit`	`integer`	No	Page size (1-200). Omit for the server default.

Review agent approvals

Changes data

atlas_agents_approvals_review

Approve or reject one pending approval. Owner/admin only. Both decisions are terminal and append an audit row; idempotent per (id, decision).

Scope: workspace:manage
Calls: POST /agent-governance/approvals/{id}/decide
Example prompt: Approve agent approval ap_42 - looks correct.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Approval row id.
`approve`	`boolean`	Yes	true = APPROVED, false = REJECTED. Both decisions are terminal.
`note`	`string`	No	Optional reviewer note - appended to the audit row.

Get agent approvals risk posture

Read only

atlas_agents_approvals_risk_posture_get

Compute the current approval-queue risk posture: aging buckets, source mix, action-kind mix, escalation candidates.

Scope: workspace:read
Calls: GET /agent-governance/approvals/risk-posture
Example prompt: What does the approval queue risk look like right now?

Arguments

This tool takes no arguments.

Send agent approvals SLA digest

Destructive

atlas_agents_approvals_sla_digest_send

Fan out the approval SLA digest to owners / admins right now. Owner/admin only. Idempotent on retry inside the digest window.

Scope: workspace:manage
Calls: POST /agent-governance/approvals/sla-digest
Example prompt: Fire off the approval SLA digest now.

Arguments

This tool takes no arguments.

Get agent execution readiness

Read only

atlas_agents_execution_readiness_get

Snapshot of agent-execution readiness: provider keys configured, kill-switch states, pending approvals at the SLA edge, replayable backlog. Read-only.

Scope: workspace:read
Calls: GET /agent-governance/execution-readiness
Example prompt: Are agents ready to run? Any provider keys missing?

Arguments

This tool takes no arguments.

Delete agent memory

Destructive

atlas_agents_memory_delete

Delete one agent memory row by id. 204 on success. Idempotent: a second delete on the same id returns 404 not 500.

Scope: workspace:manage
Calls: DELETE /agent-governance/memory/{id}
Example prompt: Forget the stale holiday memory mem_abc.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Agent governance row id (template / memory / approval).

List agent memory

Read only

atlas_agents_memory_list

List agent memory rows in the calling tenant, one page at a time. Filter by agentId and / or scope (tenant / agent / user). Expired rows are filtered out server-side. Returns { items, nextCursor }; pass nextCursor back as cursor for the next page.

Scope: workspace:read
Calls: GET /agent-governance/memory
Example prompt: List every tenant-scoped memory the planner agent has stored.

Arguments

Field	Type	Required	Description
`agentId`	`string`	No	Filter to a single agent id. Omit to list every memory row in the tenant.
`scope`	`"tenant" \| "agent" \| "user"`	No	Filter to tenant / agent / user scoped memory.
`limit`	`integer`	No	How many rows to return, 1 to 100. Defaults to 50.
`cursor`	`string`	No	The nextCursor of the previous page. Leave it out for the first page.
`sort`	`"updatedAt" \| "createdAt" \| "key"`	No	The field to sort on: updatedAt (the default), createdAt or key.
`order`	`"asc" \| "desc"`	No	asc or desc. Newest first (desc) by default.

Set agent memory

Changes data

atlas_agents_memory_set

Upsert a single agent memory row (keyed on key + scope + agentId + userId). Pass expiresAt to TTL the row. Idempotent on retry.

Scope: workspace:manage
Calls: POST /agent-governance/memory
Example prompt: Remember the office holiday list for the planner agent.

Arguments

Field	Type	Required	Description
`key`	`string`	Yes	Stable memory key. Upsert is keyed on (tenant, key, scope, agentId, userId).
`value`	`object`	Yes	Arbitrary JSON payload stored against the key.
`scope`	`"tenant" \| "agent" \| "user"`	No	Visibility scope. tenant=shared, agent=per-agent, user=per-caller.
`agentId`	`string \| null`	No	Required when scope=agent.
`userId`	`string \| null`	No	Required when scope=user.
`expiresAt`	`string (date-time) \| null`	No	Optional ISO timestamp after which the row is treated as expired.

Update agent templates governance

Changes data

atlas_agents_templates_governance_update

Update a template kill-switch and / or cost cap. Owner/admin only. The reason string is required and lands in the governance audit log.

Scope: workspace:manage
Calls: PATCH /agent-governance/templates/{id}
Example prompt: Disable the auto-reply template - investigating a refund regression.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Governed template id.
`disabled`	`boolean`	No	Flip the template kill-switch on or off. Disabled templates 409 on run.
`costCapUsd`	`number`	No	Per-run cost ceiling in USD (0 disables the gate).
`reason`	`string`	Yes	Required audit reason - lands in the AgentTemplate.governanceLog row.

List agent templates

Read only

atlas_agents_templates_list

List every governed agent template visible to the tenant - kill-switch state, per-run cost cap, and the last governance audit summary. Read-only.

Scope: workspace:read
Calls: GET /agent-governance/templates
Example prompt: Show me every governed agent template and whether any are disabled.

Arguments

This tool takes no arguments.

Run agent templates

Changes data

atlas_agents_templates_run

Queue a governed template for execution. Honours the template kill-switch and per-run cost cap server-side; returns the runId + initial status snapshot.

Scope: workspace:manage
Calls: POST /agent-governance/templates/{id}/run
Example prompt: Run the weekly customer-health digest template now.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Agent governance row id (template / memory / approval).

Agents

Every Agents tool, with the scope it needs, its arguments and the API operation it calls.

15 tools

List agent activity

Read only

atlas_agents_activity_list

Scope: workspace:read
Calls: GET /agent-governance/activity
Example prompt: What has the planner agent been doing this morning?

Arguments

Field	Type	Required	Description
`agentId`	`string`	No	Filter to a single agent id.
`status`	`"queued" \| "running" \| "completed" \| "failed" \| "approval-needed" \| "max-steps-exceeded"`	No	Filter to a single execution status.
`limit`	`integer`	No	How many rows to return, 1 to 100. Defaults to 50.
`cursor`	`string`	No	The nextCursor of the previous page. Leave it out for the first page.
`order`	`"asc" \| "desc"`	No	asc or desc. Newest first (desc) by default.

Create agent approvals

Changes data

atlas_agents_approvals_create

Create a pending approval (planner / governed template / replay drill source). Owner/admin only; idempotent on retry.

Scope: workspace:manage
Calls: POST /agent-governance/approvals
Example prompt: Open an approval for the planner to send the renewal email.

Arguments

Field	Type	Required	Description
`agentId`	`string`	Yes	Agent id requesting the approval.
`runId`	`string`	No	Optional originating run id (links the row back to a planner trace).
`source`	`"agent_planner" \| "governed_template" \| "system_replay_drill"`	Yes	Where the approval was raised from.
`actionKind`	`string`	Yes	Domain action being approved (e.g. send_email, mutate_crm_deal).
`summary`	`string`	Yes	Human-readable summary surfaced to the reviewer.
`payload`	`object`	No	Action payload the approval gates - replayed verbatim if approved.

List agent approvals

Read only

atlas_agents_approvals_list

Scope: workspace:read
Calls: GET /agent-governance/approvals
Example prompt: What pending agent approvals are waiting on me?

Arguments

Field	Type	Required	Description
`status`	`"PENDING" \| "APPROVED" \| "REJECTED"`	No	Filter to a single approval state.
`agentId`	`string`	No	Filter to a single agent id.
`limit`	`integer`	No	How many rows to return, 1 to 100. Defaults to 50.
`cursor`	`string`	No	The nextCursor of the previous page. Leave it out for the first page.
`order`	`"asc" \| "desc"`	No	asc or desc. Newest first (desc) by default.

Replay agent approvals

Changes data

atlas_agents_approvals_replay

Replay an APPROVED approval payload as a new agent run. Owner/admin only. Optional runId / reason ride along on the new replay record.

Scope: workspace:manage
Calls: POST /agent-governance/approvals/{id}/replay
Example prompt: Replay approval ap_42 as a drill.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Approved approval id to replay.
`runId`	`string`	No	Optional new run id to associate with the replay attempt.
`reason`	`string`	No	Optional replay rationale (recorded on the run row).

List agent approvals replayable

Read only

atlas_agents_approvals_replayable_list

List APPROVED approvals that are still safely replayable (within retention window, payload still resolvable). Powers the replay-drill picker.

Scope: workspace:read
Calls: GET /agent-governance/replayable-approvals
Example prompt: Show me the last 50 approvals I can replay.

Arguments

Field	Type	Required	Description
`limit`	`integer`	No	Page size (1-200). Omit for the server default.

Review agent approvals

Changes data

atlas_agents_approvals_review

Approve or reject one pending approval. Owner/admin only. Both decisions are terminal and append an audit row; idempotent per (id, decision).

Scope: workspace:manage
Calls: POST /agent-governance/approvals/{id}/decide
Example prompt: Approve agent approval ap_42 - looks correct.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Approval row id.
`approve`	`boolean`	Yes	true = APPROVED, false = REJECTED. Both decisions are terminal.
`note`	`string`	No	Optional reviewer note - appended to the audit row.

Get agent approvals risk posture

Read only

atlas_agents_approvals_risk_posture_get

Compute the current approval-queue risk posture: aging buckets, source mix, action-kind mix, escalation candidates.

Scope: workspace:read
Calls: GET /agent-governance/approvals/risk-posture
Example prompt: What does the approval queue risk look like right now?

Arguments

This tool takes no arguments.

Send agent approvals SLA digest

Destructive

atlas_agents_approvals_sla_digest_send

Fan out the approval SLA digest to owners / admins right now. Owner/admin only. Idempotent on retry inside the digest window.

Scope: workspace:manage
Calls: POST /agent-governance/approvals/sla-digest
Example prompt: Fire off the approval SLA digest now.

Arguments

This tool takes no arguments.

Get agent execution readiness

Read only

atlas_agents_execution_readiness_get

Snapshot of agent-execution readiness: provider keys configured, kill-switch states, pending approvals at the SLA edge, replayable backlog. Read-only.

Scope: workspace:read
Calls: GET /agent-governance/execution-readiness
Example prompt: Are agents ready to run? Any provider keys missing?

Arguments

This tool takes no arguments.

Delete agent memory

Destructive

atlas_agents_memory_delete

Delete one agent memory row by id. 204 on success. Idempotent: a second delete on the same id returns 404 not 500.

Scope: workspace:manage
Calls: DELETE /agent-governance/memory/{id}
Example prompt: Forget the stale holiday memory mem_abc.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Agent governance row id (template / memory / approval).

List agent memory

Read only

atlas_agents_memory_list

Scope: workspace:read
Calls: GET /agent-governance/memory
Example prompt: List every tenant-scoped memory the planner agent has stored.

Arguments

Field	Type	Required	Description
`agentId`	`string`	No	Filter to a single agent id. Omit to list every memory row in the tenant.
`scope`	`"tenant" \| "agent" \| "user"`	No	Filter to tenant / agent / user scoped memory.
`limit`	`integer`	No	How many rows to return, 1 to 100. Defaults to 50.
`cursor`	`string`	No	The nextCursor of the previous page. Leave it out for the first page.
`sort`	`"updatedAt" \| "createdAt" \| "key"`	No	The field to sort on: updatedAt (the default), createdAt or key.
`order`	`"asc" \| "desc"`	No	asc or desc. Newest first (desc) by default.

Set agent memory

Changes data

atlas_agents_memory_set

Upsert a single agent memory row (keyed on key + scope + agentId + userId). Pass expiresAt to TTL the row. Idempotent on retry.

Scope: workspace:manage
Calls: POST /agent-governance/memory
Example prompt: Remember the office holiday list for the planner agent.

Arguments

Field	Type	Required	Description
`key`	`string`	Yes	Stable memory key. Upsert is keyed on (tenant, key, scope, agentId, userId).
`value`	`object`	Yes	Arbitrary JSON payload stored against the key.
`scope`	`"tenant" \| "agent" \| "user"`	No	Visibility scope. tenant=shared, agent=per-agent, user=per-caller.
`agentId`	`string \| null`	No	Required when scope=agent.
`userId`	`string \| null`	No	Required when scope=user.
`expiresAt`	`string (date-time) \| null`	No	Optional ISO timestamp after which the row is treated as expired.

Update agent templates governance

Changes data

atlas_agents_templates_governance_update

Update a template kill-switch and / or cost cap. Owner/admin only. The reason string is required and lands in the governance audit log.

Scope: workspace:manage
Calls: PATCH /agent-governance/templates/{id}
Example prompt: Disable the auto-reply template - investigating a refund regression.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Governed template id.
`disabled`	`boolean`	No	Flip the template kill-switch on or off. Disabled templates 409 on run.
`costCapUsd`	`number`	No	Per-run cost ceiling in USD (0 disables the gate).
`reason`	`string`	Yes	Required audit reason - lands in the AgentTemplate.governanceLog row.

List agent templates

Read only

atlas_agents_templates_list

List every governed agent template visible to the tenant - kill-switch state, per-run cost cap, and the last governance audit summary. Read-only.

Scope: workspace:read
Calls: GET /agent-governance/templates
Example prompt: Show me every governed agent template and whether any are disabled.

Arguments

This tool takes no arguments.

Run agent templates

Changes data

atlas_agents_templates_run

Queue a governed template for execution. Honours the template kill-switch and per-run cost cap server-side; returns the runId + initial status snapshot.

Scope: workspace:manage
Calls: POST /agent-governance/templates/{id}/run
Example prompt: Run the weekly customer-health digest template now.

Arguments

Field	Type	Required	Description
`id`	`string`	Yes	Agent governance row id (template / memory / approval).