agent-graph-mcp 0.2.5

MCP server for graph-orchestrated LLM workflows — 25 typed tools, daemon/proxy architecture, checkpoint/resume, HITL approvals, HMAC receipts
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127

{
  "description": "MCP server for graph-orchestrated LLM workflows — 25 typed tools, daemon/proxy architecture, checkpoint/resume, human-in-the-loop approvals, HMAC-authenticated execution receipts, and SQLite-backed durable persistence. Compile declarative workflow specs to the ri-agent-graph runtime engine. Supports parallel fan-out with JoinSet-backed concurrency, conditional routing, 5 join modes, state transforms, subgraph composition, and 5 built-in templates (council_deliberation, parallel_council, plan_critique_refine, analysis_pipeline, classifier_router). Ships as a daemon (agent-graph-mcpd) with file-lock ownership and crash recovery, plus a stateless stdin/stdout proxy that bridges MCP clients to the daemon over a Unix socket. All integrity-sensitive operations use HMAC-SHA256. Built in Rust on Tokio + rmcp.",
  "homepage": "https://github.com/RecursiveIntell/agent-graph-mcp",
  "identifier": "recursiveintell-agent-graph-mcp",
  "name": "Agent Graph MCP",
  "tags": [
    "mcp",
    "agent-orchestration",
    "graph-workflow",
    "llm-pipeline",
    "rust",
    "langgraph",
    "parallel-execution",
    "checkpointing",
    "human-in-the-loop",
    "workflow-engine",
    "fan-out",
    "template-engine",
    "council",
    "sqlite",
    "deterministic-resume"
  ],
  "version": "1.0.2",
  "capabilities": {
    "tools": true,
    "prompts": false,
    "resources": false
  },
  "tools": [
    {
      "name": "graph_create",
      "description": "Create, validate, or delete a graph-orchestrated workflow from a JSON spec. Supports template instantiation and idempotency keys.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "action": { "type": "string", "description": "Action: 'create', 'validate', or 'delete'" },
          "spec": { "type": "object", "description": "JSON graph specification" },
          "template": { "type": "object", "description": "Template instantiation: { 'id': 'council', 'name': 'my-council' }" },
          "graph_id": { "type": "string", "description": "Graph ID for delete action" },
          "idempotency_key": { "type": "string", "description": "Optional idempotency key" },
          "overwrite": { "type": "boolean", "description": "Overwrite existing graph" }
        }
      }
    },
    {
      "name": "graph_execute",
      "description": "Execute a registered graph. Sync mode blocks until completion; async mode returns immediately with a run_id.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "graph_id": { "type": "string", "description": "The graph ID to execute" },
          "input": { "type": "object", "description": "Input value to pass to the graph's entry node" },
          "mode": { "type": "string", "description": "'sync' blocks until completion, 'async' returns immediately" },
          "idempotency_key": { "type": "string", "description": "Optional idempotency key" }
        },
        "required": ["graph_id"]
      }
    },
    {
      "name": "graph_run_start",
      "description": "Start an async graph run. Returns run_id immediately. Supports optional max_wall_clock_ms and max_nodes budgets.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "graph_id": { "type": "string", "description": "Graph ID to execute" },
          "input": { "type": "object", "description": "Input value" },
          "budgets": { "type": "object", "description": "Optional budgets with max_wall_clock_ms or max_nodes" },
          "checkpoint": { "type": "boolean", "description": "When true, persist a pre-execution checkpoint" }
        },
        "required": ["graph_id"]
      }
    },
    {
      "name": "graph_run_wait",
      "description": "Wait for an async run to complete, with optional timeout.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "run_id": { "type": "string" },
          "timeout_ms": { "type": "integer" }
        },
        "required": ["run_id"]
      }
    },
    {
      "name": "graph_run_receipt",
      "description": "Fetch the canonical HMAC-SHA256 authenticated execution receipt for a run.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "run_id": { "type": "string" }
        },
        "required": ["run_id"]
      }
    },
    {
      "name": "graph_template_list",
      "description": "List available built-in graph templates: council_deliberation, parallel_council, plan_critique_refine, analysis_pipeline, classifier_router.",
      "inputSchema": {
        "type": "object",
        "properties": {}
      }
    },
    {
      "name": "graph_inspect",
      "description": "Get a graph's full topology: nodes, edges, Mermaid diagram, and topology hash.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "graph_id": { "type": "string" }
        },
        "required": ["graph_id"]
      }
    },
    {
      "name": "graph_list",
      "description": "List all registered graphs with metadata (name, node count, edge count, version).",
      "inputSchema": {
        "type": "object",
        "properties": {
          "limit": { "type": "integer" },
          "query": { "type": "string" }
        }
      }
    }
  ]
}