rmems/mcp-tool-use-episodes

MCP Tool-Use Episodes

Purpose

Multi-turn MCP/tool-calling episodes against real tool schemas, including wrong-argument errors, schema violations, and the agent's recovery, rather than single clean calls.

Record schema

JSONL, one record per line. Fields:

record_id, family, tools[{name, input_schema}], messages[{role, content, tool_calls[{name, arguments}], tool_result{ok, error}}], recovery{error_turns, recovered}, result{task_completed, verified_by}, provenance{produced_at, actors, lineage}

Illustrative example (EXAMPLE ONLY, not real data):

{
  "record_id": "mcp-r01-000001",
  "family": "mcp-tool-use-episode",
  "tools": [
    {
      "name": "search_issues",
      "input_schema": {
        "query": "string",
        "limit": "integer"
      }
    }
  ],
  "messages": [
    {
      "role": "user",
      "content": "Find open bugs about login."
    },
    {
      "role": "assistant",
      "tool_calls": [
        {
          "name": "search_issues",
          "arguments": {
            "query": "login bug",
            "limit": "ten"
          }
        }
      ]
    },
    {
      "role": "tool",
      "tool_result": {
        "ok": false,
        "error": "limit: expected integer"
      }
    },
    {
      "role": "assistant",
      "tool_calls": [
        {
          "name": "search_issues",
          "arguments": {
            "query": "login bug",
            "limit": 10
          }
        }
      ]
    },
    {
      "role": "tool",
      "tool_result": {
        "ok": true
      }
    }
  ],
  "recovery": {
    "error_turns": 1,
    "recovered": true
  },
  "result": {
    "task_completed": true,
    "verified_by": "replay"
  }
}

Generation

• Open-weight local models only, run via Ollama: gemma4:12b (writer/solver), qwen3-embedding:8b (embedding dedup), FrogNano-4B (Apache-2.0). No hosted frontier models.
• Produced via rmems/synthetic-factory, with provenance gates (every record names its actors, model digests, and seeds) and execution-evidence gates (claims are re-derived from measured execution, not trusted from model transcripts).

Intended use

Post-training of open-weight coding/agent models: supervised fine-tuning (SFT) on trajectories and preference tuning (DPO). Not intended for pre-training or for evaluating hosted models.

Status

17 records (train: 17), generated deterministically by the rmems/synthetic-factory CLI at commit 5e3b048. Every record passed the factory's replay gate; each carries its lane, generate/replay commands, seed and source record id under provenance.lineage. No LLM was used. Field names follow the schema below, with extra fields where the factory emits more detail.

Limitations

• Current records come from deterministic scripted/programmatic generators (no LLM); trajectories may be narrower and simpler than real-world agent behavior.
• Task sources (e.g. public repos) are likely present in base-model pretraining; contamination checks and program-level eval splits are planned but not yet validated.
• English only. Schema may still change in later releases.

Known limitations (this release)

• Small: only 4 tasks in the tool-world-v1 MCP factory; seeds mostly reproduce identical episodes (deduplicated).
• Trajectories come from a scripted policy (tool-world-scripted-policy), not a model.
• recovery.error_turns counts only JSON-RPC errors and isError results; injected faults such as an empty first tools page are not counted there, so use recovery.faults_fired.
• One give_up_on_fault-variant episode is included because it still succeeded and passed all hidden predicates.
• Tool-call args/arguments are stored as JSON-encoded strings (some episodes deliberately pass wrongly typed arguments, which would otherwise break a fixed column type); json.loads them to recover the original object.