Oriel — AI & technology worth noticing, every day.

ToolSpace: Scaling Tool-Centric Environment Synthesis for Long-Horizon Agents

Paper · https://arxiv.org/abs/2611.05036

Large language models are deployed as long-horizon agents that compose diverse tools across MCP services into executable workflows. Existing MCP environment-scaling methods are data-centric: they first synthesize a database and scenario, then derive tools and tasks from them. As a result, tool interfaces may not match real ones, tool compositions are not grounded in execution, and tasks rarely span more than one environment. We propose ToolSpace, a tool-centric framework that reverses this order. ToolSpace mines data, state, and retry dependencies from adjacent calls in execution traces of real MCP services and open agent corpora, using deterministic rules rather than LLM judgment, and clusters the resulting weighted tool-composition graph into environments. Databases are synthesized only for state-mutating environments, while query and stateless tools replay recorded returns. ToolSpace then performs state-aware random walks with rollback and decompiles each chain into a natural-language task with assertion-based checks, keeping only samples that replay successfully from the initial state. Applied to AutomationBench and tau3-Bench Banking, the pipeline produces harder Pro versions on which 7 frontier models drop by up to 38.2 points. Across seven long-horizon tool-use benchmarks, ToolSpace raises Qwen3.6-35B-A3B from 23.0 to 41.0 average score (43.2 with RL) and GLM-5.2 from 44.6 to 56.6, on par with Claude Opus 5. Ablations show that execution-witnessed edges clearly outperform LLM-judged ones, and performance keeps improving with data and environment scale.

First seen by Oriel on .

AI summary

ToolSpace is a tool-centric framework for synthesizing executable, state-aware environments and harder benchmarks for long-horizon agents using execution traces from real MCP services.

  • Existing MCP environment-scaling methods synthesize databases and scenarios first, causing tool-interface mismatches, ungrounded compositions, and mostly single-environment tasks.
  • ToolSpace deterministically mines data, state, and retry dependencies from adjacent calls, then clusters a weighted tool-composition graph into environments.
  • State-mutating environments receive synthesized databases, while query and stateless tools replay recorded returns during state-aware random walks with rollback.
  • Applied to AutomationBench and tau3-Bench Banking, ToolSpace produced Pro versions where seven frontier models dropped by up to 38.2 points.
  • Across seven benchmarks, Qwen3.6-35B-A3B improved from 23.0 to 41.0 average score, or 43.2 with RL, while GLM-5.2 reached 56.6.

#mcp #long-horizon agents #tool use #benchmarking

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 45, latest observation ; +12 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 4, latest observation ; +3 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2611.05036https://arxiv.org/abs/2611.05036 Read discussion