MCP Tool Reference
The Foxhound MCP server exposes 36 tools, grouped by capability.
Trace Querying
foxhound_search_traces
Search traces by agent name, time range, and status. Returns a summary list of matching traces.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
agent_name | string | No | Filter by agent ID/name |
from | string | No | Start time (ISO 8601 or epoch ms) |
to | string | No | End time (ISO 8601 or epoch ms) |
limit | number | No | Max results (default 20, max 100) |
Example prompts:
Show me all traces for agent billing-bot in the last hour
Find error traces from the past 24 hours
foxhound_get_trace
Get the full trace with its complete span tree. Use this to inspect what happened during an agent run.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace ID to retrieve |
Example prompts:
Get trace abc-123 and show me the span tree
What happened in trace def-456?
foxhound_replay_span
Reconstruct the full agent state at the moment a specific span began — including LLM context, tool inputs, and memory. Requires Pro plan.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace ID containing the span |
span_id | string | Yes | The span ID to replay |
Example prompts:
Replay span xyz in trace abc-123 — what was the agent's context?
foxhound_diff_runs
Compare two agent runs side-by-side and surface divergence points. Useful for debugging regressions. Requires Pro plan.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id_a | string | Yes | First trace/run ID |
trace_id_b | string | Yes | Second trace/run ID |
Example prompts:
Compare runs abc and def — why did the second one fail?
Analysis
foxhound_get_anomalies
Surface behavioral anomalies in recent traces for an agent — unusually slow spans, error spikes, or unexpected tool usage patterns.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
agent_name | string | Yes | Agent ID/name to analyze |
hours | number | No | Lookback window in hours (default 24, max 168) |
Example prompts:
Any anomalies for billing-bot in the last 12 hours?
Show me error spikes for the onboarding agent
foxhound_explain_failure
Analyze a failed trace and produce a human-readable explanation of what went wrong, including the error chain and likely root cause.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace ID of the failed run |
foxhound_suggest_fix
Suggest code-level fixes for a failed trace, based on the error chain and span context.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace ID to suggest fixes for |
Scoring
foxhound_score_trace
Manually score a trace on one or more dimensions (e.g. helpfulness, accuracy, safety).
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace to score |
dimension | string | Yes | Scoring dimension (e.g. helpfulness) |
score | number | Yes | Score value (0.0–1.0) |
rationale | string | No | Optional explanation for the score |
foxhound_get_trace_scores
Retrieve all scores attached to a trace across all dimensions and evaluators.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace to retrieve scores for |
Evaluators
foxhound_list_evaluators
List all evaluators configured for your organization.
Parameters: None
foxhound_run_evaluator
Trigger async evaluator runs for one or more traces. Evaluator runs are async — use foxhound_get_evaluator_run to check status and results.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
evaluator_id | string | Yes | The evaluator to run |
trace_ids | string[] | Yes | List of trace IDs to evaluate |
foxhound_get_evaluator_run
Check the status and results of an async evaluator run.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
run_id | string | Yes | The evaluator run ID |
Datasets
foxhound_list_datasets
List all datasets in your organization.
Parameters: None
foxhound_add_trace_to_dataset
Add a trace to a dataset for evaluation or fine-tuning purposes.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
trace_id | string | Yes | The trace to add |
dataset_id | string | Yes | The target dataset |
foxhound_curate_dataset
Curate a dataset by filtering traces based on quality criteria.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
dataset_id | string | Yes | The dataset to curate |
Alerts
foxhound_list_alert_rules
List all alert rules configured for your organization.
Parameters: None
foxhound_create_alert_rule
Create a new alert rule that routes events to a notification channel. This is a write operation.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
event_type | enum | Yes | agent_failure, anomaly_detected, cost_spike, compliance_violation |
min_severity | enum | No | critical, high, medium, low (default: high) |
channel_id | string | Yes | The notification channel ID to route alerts to |
foxhound_delete_alert_rule
Delete an alert rule by ID. Set confirm=true to execute. Without confirmation, returns a preview of what will be deleted.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
rule_id | string | Yes | The alert rule ID to delete |
confirm | boolean | No | Set to true to confirm deletion. Omit to preview |
foxhound_list_channels
List all notification channels (e.g. Slack webhooks) configured for your organization.
Parameters: None
foxhound_create_channel
Create a new Slack notification channel. This is a write operation.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
name | string | Yes | A human-readable name for the channel |
webhook_url | string | Yes | The Slack incoming webhook URL |
slack_channel | string | No | Optional Slack channel override |
foxhound_test_channel
Send a test alert through a notification channel to verify it works.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
channel_id | string | Yes | The channel ID to test |
foxhound_delete_channel
Delete a notification channel by ID. This may also delete associated alert rules. Set confirm=true to execute.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
channel_id | string | Yes | The channel ID to delete |
confirm | boolean | No | Set to true to confirm deletion. Omit to preview |
Keys & Budget
foxhound_list_api_keys
List active API keys for your organization. Keys are masked — only the prefix is shown.
Parameters: None
foxhound_create_api_key
Create a new API key. For security, the plaintext key is not returned through MCP — use the CLI (foxhound keys create) or dashboard to retrieve it.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
name | string | Yes | A human-readable name for the key |
foxhound_revoke_api_key
Revoke an API key by ID. The key will immediately stop working. Set confirm=true to execute.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
key_id | string | Yes | The API key ID to revoke |
confirm | boolean | No | Set to true to confirm revocation. Omit to preview |
foxhound_get_agent_budget
Get the cost budget configuration and current spend status for a specific agent.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
agentId | string | Yes | The agent ID to retrieve budget for |
foxhound_get_cost_summary
Get token usage and cost breakdown. Shows current billing period span usage and limits.
Parameters: None
SLA & Baselines
foxhound_check_sla_status
Check SLA targets and compliance status for a specific agent, including p95 duration and success rate.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
agentId | string | Yes | The agent ID to check SLA status for |
foxhound_detect_regression
Compare two versions of an agent and detect span-level regressions — missing or newly added spans between versions.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
agentId | string | Yes | The agent ID to analyze |
versionA | string | Yes | The baseline version (before) |
versionB | string | Yes | The comparison version (after) |
foxhound_list_baselines
List all stored baseline snapshots for an agent, showing version, sample size, and creation date.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
agentId | string | Yes | The agent ID to list baselines for |
foxhound_status
Check the health and connectivity of the Foxhound server.
Parameters: None
Prompt Management
foxhound_list_prompts
List all prompt templates in the registry. Shows prompt names, IDs, and timestamps. Requires Pro plan.
Parameters: None
foxhound_get_prompt
Resolve a prompt by name and label (defaults to "production"). Returns the prompt content, model, config, and version number. Use this to inspect what prompt version is currently active.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
name | string | Yes | The prompt name (e.g. "support-agent") |
label | string | No | The label to resolve (default: "production"). Common labels: production, staging, canary |
Example prompts:
What's the current production prompt for support-agent?
Show me the staging version of billing-classifier
foxhound_list_prompt_versions
List all versions of a prompt, including their labels. Use the prompt ID (not name) from foxhound_list_prompts.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt_id | string | Yes | The prompt ID (e.g. "pmt_...") |
foxhound_create_prompt
Create a new prompt in the registry. The name must be alphanumeric with hyphens/underscores (no spaces). Requires Pro plan. This is a write operation.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
name | string | Yes | Prompt name (alphanumeric, hyphens, underscores only) |
foxhound_create_prompt_version
Add a new version to an existing prompt. The version number auto-increments. Requires Pro plan. This is a write operation.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt_id | string | Yes | The prompt ID (e.g. "pmt_...") |
content | string | Yes | The prompt template content |
model | string | No | Optional model recommendation (e.g. "gpt-4o", "claude-sonnet-4-20250514") |
foxhound_set_prompt_label
Set a label (e.g. "production", "staging") on a specific prompt version. If the label already exists on another version of the same prompt, it is moved. This is a write operation.
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt_id | string | Yes | The prompt ID (e.g. "pmt_...") |
version_number | number | Yes | The version number to label |
label | string | Yes | Label name (alphanumeric, hyphens, underscores — e.g. "production", "staging") |
Example prompts:
Create a new prompt called billing-classifier
Add a version to the support-agent prompt with this content: "You are a helpful support agent..."
Set the production label on version 3 of support-agent