MCP tools
44 tools, generated from the MCP server’s own tool list. Every one only reads: no tool approves, pays, or changes a limit.
| Tool | What it returns | Role |
|---|---|---|
active_runs |
Runs in progress now (contract 1.12): | viewer and above |
agent |
One agent: | viewer and above |
agent_map |
The agent map, read-only. | viewer and above |
agents |
Your agents with live state (running, idle, failing, paused, awaiting sign-off), autonomy, last run, runs and cost this month. | viewer and above |
attention |
The attention queue, read-only: | viewer and above |
billing_summary |
This month’s model spend (direct and through AAL), the managed charge at the price card, the prepaid credit balance, spend by provider connection, and limit alerts. | viewer and above |
billing_usage |
Model spend for a month (YYYY-MM), grouped by day, model, pack, agent or connection, with the managed charge. | viewer and above |
caching |
Prompt caching by day, model, agent, pack, deployment or provider: | viewer and above |
command_history |
Handler commands, newest first, for one agent (<pack>:<card>) or the whole organisation, optionally by status. |
viewer and above |
command_status |
One handler command (pause, cap, pre-approval, approval, release) with its status history, by command id. | viewer and above |
commercial_terms |
Your commercial terms with AAL (Billing → Terms): | client_admin |
connection |
One connection by its id (cx-...): |
viewer and above |
connectors |
The connector catalogue and this organisation’s connections: | viewer and above |
deployment_profiles |
Each deployment’s profile (contract 1.13): | viewer and above |
deployment_wakes |
How long deployments took to wake (contract 1.13): | viewer and above |
evals |
Evaluation results, filtered by pack and family. | viewer and above |
evidence |
Evidence scores (contract 1.14): | viewer and above |
forecast |
Projected month-end spend from the run rate and weekday pattern, with a likely range, against the cap. | viewer and above |
health |
Latest health of each component and connector. | viewer and above |
kills |
Kill-switch stops and revivals, newest first, and how many are in force. | viewer and above |
knowledge |
Knowledge records (what was asked, read and cited), read in place from your own account. | content_viewer and above |
knowledge_base |
One knowledge base by its index id (contract 1.14): | viewer and above |
knowledge_bases |
Knowledge bases (contract 1.14): | viewer and above |
library |
The agent library: | viewer and above |
limits |
Spend limits (tenant, pack, agent, person; daily or monthly): | viewer and above |
locale |
Your effective time zone, display currency, date and number formats and week start: | viewer and above |
market_usage |
Your marketplace listings and their review state, and this month’s use of them by subscriber (a pseudonym). | client_admin |
model_picker |
Recommended models per use case, ranked by evidence, filtered by the chosen cloud and the data-residency rule, with the estimated routing saving against the top model and which models were excluded and why. | viewer and above |
models |
The models your deployments may call: | viewer and above |
preapprovals |
Standing pre-approvals: | viewer and above |
providers |
Model provider connections (direct or through AAL), their health and the models each exposes. | viewer and above |
rate_limits |
Peak-minute utilisation of requests, input tokens and output tokens per minute by deployment and model, against the provider quota and the gateway’s pacing limit, with throttles and pacing waits. | viewer and above |
recommendations |
Cheaper-model moves backed by evaluation parity, and waste flags per agent, with their evidence. | viewer and above |
refusals |
Refused model calls by reason, and screening counts. | viewer and above |
report |
The latest evidence or value report, or the one for a month (YYYY-MM). | viewer and above |
retrieval |
Retrieval as a measured outcome (contract 1.7): | viewer and above |
retrieval_savings |
Estimated saving of retrieval against context stuffing (contract 1.7, stable): | viewer and above |
run |
One run by its run_id_hash, whenever it started. | viewer and above |
run_map |
One run as a tree of its steps and sub-agents (contract 1.12), each node with tokens, cost, duration, status, model and retries; the critical path, the most expensive and the slowest node, and totals. | viewer and above |
run_targets |
Where this organisation’s agents run: | viewer and above |
runs |
Runs, newest first, filtered by pack, card, status and start time. | viewer and above |
savings |
Savings. | viewer and above |
setup |
The setup mission: | viewer and above |
usage |
Model usage, tokens and cost, grouped by model, pack, task or day. | viewer and above |