Skip to main content

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:

ParameterTypeRequiredDescription
agent_namestringNoFilter by agent ID/name
fromstringNoStart time (ISO 8601 or epoch ms)
tostringNoEnd time (ISO 8601 or epoch ms)
limitnumberNoMax 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:

ParameterTypeRequiredDescription
trace_idstringYesThe 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:

ParameterTypeRequiredDescription
trace_idstringYesThe trace ID containing the span
span_idstringYesThe 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:

ParameterTypeRequiredDescription
trace_id_astringYesFirst trace/run ID
trace_id_bstringYesSecond 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:

ParameterTypeRequiredDescription
agent_namestringYesAgent ID/name to analyze
hoursnumberNoLookback 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:

ParameterTypeRequiredDescription
trace_idstringYesThe 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:

ParameterTypeRequiredDescription
trace_idstringYesThe 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:

ParameterTypeRequiredDescription
trace_idstringYesThe trace to score
dimensionstringYesScoring dimension (e.g. helpfulness)
scorenumberYesScore value (0.0–1.0)
rationalestringNoOptional explanation for the score

foxhound_get_trace_scores

Retrieve all scores attached to a trace across all dimensions and evaluators.

Parameters:

ParameterTypeRequiredDescription
trace_idstringYesThe 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:

ParameterTypeRequiredDescription
evaluator_idstringYesThe evaluator to run
trace_idsstring[]YesList of trace IDs to evaluate

foxhound_get_evaluator_run

Check the status and results of an async evaluator run.

Parameters:

ParameterTypeRequiredDescription
run_idstringYesThe 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:

ParameterTypeRequiredDescription
trace_idstringYesThe trace to add
dataset_idstringYesThe target dataset

foxhound_curate_dataset

Curate a dataset by filtering traces based on quality criteria.

Parameters:

ParameterTypeRequiredDescription
dataset_idstringYesThe 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:

ParameterTypeRequiredDescription
event_typeenumYesagent_failure, anomaly_detected, cost_spike, compliance_violation
min_severityenumNocritical, high, medium, low (default: high)
channel_idstringYesThe 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:

ParameterTypeRequiredDescription
rule_idstringYesThe alert rule ID to delete
confirmbooleanNoSet 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:

ParameterTypeRequiredDescription
namestringYesA human-readable name for the channel
webhook_urlstringYesThe Slack incoming webhook URL
slack_channelstringNoOptional Slack channel override

foxhound_test_channel

Send a test alert through a notification channel to verify it works.

Parameters:

ParameterTypeRequiredDescription
channel_idstringYesThe 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:

ParameterTypeRequiredDescription
channel_idstringYesThe channel ID to delete
confirmbooleanNoSet 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:

ParameterTypeRequiredDescription
namestringYesA 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:

ParameterTypeRequiredDescription
key_idstringYesThe API key ID to revoke
confirmbooleanNoSet 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:

ParameterTypeRequiredDescription
agentIdstringYesThe 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:

ParameterTypeRequiredDescription
agentIdstringYesThe 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:

ParameterTypeRequiredDescription
agentIdstringYesThe agent ID to analyze
versionAstringYesThe baseline version (before)
versionBstringYesThe comparison version (after)

foxhound_list_baselines

List all stored baseline snapshots for an agent, showing version, sample size, and creation date.

Parameters:

ParameterTypeRequiredDescription
agentIdstringYesThe 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:

ParameterTypeRequiredDescription
namestringYesThe prompt name (e.g. "support-agent")
labelstringNoThe 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:

ParameterTypeRequiredDescription
prompt_idstringYesThe 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:

ParameterTypeRequiredDescription
namestringYesPrompt 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:

ParameterTypeRequiredDescription
prompt_idstringYesThe prompt ID (e.g. "pmt_...")
contentstringYesThe prompt template content
modelstringNoOptional 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:

ParameterTypeRequiredDescription
prompt_idstringYesThe prompt ID (e.g. "pmt_...")
version_numbernumberYesThe version number to label
labelstringYesLabel 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