Get Started
MCP Tools
AgenticReality exposes 15 compact MCP tools in default (compact) surface mode. Each tool is a domain facade accepting an operation field plus operation-specific parameters.
AgenticReality exposes 15 compact MCP tools in default (compact) surface mode. Each tool is a domain facade accepting an operation field plus operation-specific parameters.
Tool List
| Tool | Domain | Key Operations |
|---|---|---|
reality_workspace | Workspace management | init, sense, save, load, info, status |
reality_soul | Deployment soul + incarnation | init, summary, show, incarnation, substrate, fingerprint |
reality_memory | Incarnation memory | incarnations, wisdom, karma, recall, forget |
reality_resource | Resource body | body, pressure, capacity, cost, capabilities |
reality_layer | Reality layers | get, add, set_freshness, list, detect_simulation |
reality_anchor | Ground truth anchors | create, verify, list, delete, refresh |
reality_hallucination | Hallucination detection | check, scan, report, score |
reality_topology | Service mesh topology | map, neighbors, dependencies, observers |
reality_temporal | Temporal grounding | now, causality, timeline, anchor_time |
reality_stakes | Stakes perception | assess, risk_field, blast_radius, get |
reality_coherence | Coherence maintenance | check, report, resolve, status |
reality_context | Unified context | summary, shift_check, diff, fingerprint |
reality_ground | Claim grounding | verify, status, evidence, suggest |
reality_deployment | Deployment lifecycle | birth, rebirth, death, transition |
reality_workspace_file | .areal file management | list, inspect, compare, export |
Tool Schemas
reality_workspace
{
"operation": "sense",
"domains": ["deployment", "resources", "topology"]
}| Operation | Parameters | Description |
|---|---|---|
init | — | Initialize workspace |
sense | domains? (array) | Sense all or selected domains |
save | — | Persist to .areal |
load | path | Load from .areal |
info | — | Workspace metadata |
status | — | Current sensing status |
reality_anchor
{
"operation": "create",
"source": "health-check",
"claim": "database is healthy",
"evidence": "response time 12ms, status 200"
}| Operation | Parameters | Description |
|---|---|---|
create | source, claim, evidence | Create a ground-truth anchor |
verify | claim | Verify claim against anchors |
list | — | List all anchors |
delete | anchor_id | Delete an anchor |
refresh | anchor_id | Refresh anchor from source |
reality_ground
{
"operation": "verify",
"claim": "the database handles 10,000 RPS"
}Returns:
{
"status": "UNGROUNDED",
"confidence": 0.0,
"matching_anchors": [],
"risk": "HIGH",
"message": "No anchors match this claim. Hallucination risk is high."
}reality_stakes
{ "operation": "assess" }Returns:
{
"level": "HIGH",
"consequence": "financial transactions, 50000 active users",
"risk_field": ["external-payments", "auth-service", "user-data"],
"blast_radius": 50000,
"recommendation": "Apply extra caution. Prefer reversible actions."
}reality_coherence
{ "operation": "check" }Returns:
{
"coherent": true,
"violations": 0,
"contradictions": 0,
"domains_checked": 7,
"score": 0.98
}reality_hallucination
{
"operation": "check",
"claim": "memory usage is under 50%"
}Returns:
{
"claim": "memory usage is under 50%",
"risk": "LOW",
"confidence": 0.91,
"grounding_evidence": ["resource anchor: mem=43%, timestamp: 30s ago"]
}Error Handling
All tools return a standard error shape on failure:
{
"error": "storage_not_found",
"message": "No .areal file found. Run reality_workspace with operation init first.",
"suggestion": "reality_workspace { \"operation\": \"init\" }"
}Tools never throw JSON-RPC errors for execution failures. Only protocol-level failures (unknown tool, malformed JSON) produce JSON-RPC error responses.
Full Surface Mode
Set AREAL_MCP_TOOL_SURFACE=full to expose all 78 granular operations as individual tools. This is intended for development and debugging only.