{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WP3SBP3E7JGJRLX474SXYLPM7T","short_pith_number":"pith:WP3SBP3E","schema_version":"1.0","canonical_sha256":"b3f720bf64fa4c98aefcff257c2decfcd9c3e260d858ce594aaee0f985776512","source":{"kind":"arxiv","id":"2505.03275","version":1},"attestation_state":"computed","paper":{"title":"RAG-MCP: Mitigating Prompt Bloat in LLM Tool Selection via Retrieval-Augmented Generation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Qiyao Sun, Tiantian Gan","submitted_at":"2025-05-06T08:05:35Z","abstract_excerpt":"Large language models (LLMs) struggle to effectively utilize a growing number of external tools, such as those defined by the Model Context Protocol (MCP)\\cite{IntroducingMCP}, due to prompt bloat and selection complexity. We introduce RAG-MCP, a Retrieval-Augmented Generation framework that overcomes this challenge by offloading tool discovery. RAG-MCP uses semantic retrieval to identify the most relevant MCP(s) for a given query from an external index before engaging the LLM. Only the selected tool descriptions are passed to the model, drastically reducing prompt size and simplifying decisio"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2505.03275","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-06T08:05:35Z","cross_cats_sorted":["cs.SE"],"title_canon_sha256":"bad09253c4f5de203c6fbfd82d92832f69748e6216411bf00af386f249b8ee22","abstract_canon_sha256":"11e85f1314fbfa6e32a20bfaee0b8d9d806ffe6b2a8d217177a35df5d23322c8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:59:13.225926Z","signature_b64":"mWrb5uWfhfbuHnmoiEPGJQPFHHQAKKLTj09WLKTZMRPCfBMaeQiN2BtE/xKcFQ5+9KVErEKmzvmu6OEGC84kDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b3f720bf64fa4c98aefcff257c2decfcd9c3e260d858ce594aaee0f985776512","last_reissued_at":"2026-07-05T10:59:13.225454Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:59:13.225454Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RAG-MCP: Mitigating Prompt Bloat in LLM Tool Selection via Retrieval-Augmented Generation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Qiyao Sun, Tiantian Gan","submitted_at":"2025-05-06T08:05:35Z","abstract_excerpt":"Large language models (LLMs) struggle to effectively utilize a growing number of external tools, such as those defined by the Model Context Protocol (MCP)\\cite{IntroducingMCP}, due to prompt bloat and selection complexity. We introduce RAG-MCP, a Retrieval-Augmented Generation framework that overcomes this challenge by offloading tool discovery. RAG-MCP uses semantic retrieval to identify the most relevant MCP(s) for a given query from an external index before engaging the LLM. Only the selected tool descriptions are passed to the model, drastically reducing prompt size and simplifying decisio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.03275","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2505.03275/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2505.03275","created_at":"2026-07-05T10:59:13.225512+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.03275v1","created_at":"2026-07-05T10:59:13.225512+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.03275","created_at":"2026-07-05T10:59:13.225512+00:00"},{"alias_kind":"pith_short_12","alias_value":"WP3SBP3E7JGJ","created_at":"2026-07-05T10:59:13.225512+00:00"},{"alias_kind":"pith_short_16","alias_value":"WP3SBP3E7JGJRLX4","created_at":"2026-07-05T10:59:13.225512+00:00"},{"alias_kind":"pith_short_8","alias_value":"WP3SBP3E","created_at":"2026-07-05T10:59:13.225512+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05936","citing_title":"Mitigating Errors in LLM-Generated Web API Invocations via Retrieval-Augmented Generation and Constrained Decoding","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2606.17519","citing_title":"Scaling Enterprise Agent Routing: Degradation, Diagnosis, and Recovery","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06284","citing_title":"ToolChoiceConfusion: Causal Minimal Tool Filtering for Reliable LLM Agents","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24050","citing_title":"More Skills, Worse Agents? Skill Shadowing Degrades Performance When Expanding Skill Libraries","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30317","citing_title":"MCP Server Architecture Patterns for LLM-Integrated Applications","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24660","citing_title":"How Many Tools Should an LLM Agent See? A Chance-Corrected Answer","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10787","citing_title":"ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10787","citing_title":"ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T","json":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T.json","graph_json":"https://pith.science/api/pith-number/WP3SBP3E7JGJRLX474SXYLPM7T/graph.json","events_json":"https://pith.science/api/pith-number/WP3SBP3E7JGJRLX474SXYLPM7T/events.json","paper":"https://pith.science/paper/WP3SBP3E"},"agent_actions":{"view_html":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T","download_json":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T.json","view_paper":"https://pith.science/paper/WP3SBP3E","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.03275&json=true","fetch_graph":"https://pith.science/api/pith-number/WP3SBP3E7JGJRLX474SXYLPM7T/graph.json","fetch_events":"https://pith.science/api/pith-number/WP3SBP3E7JGJRLX474SXYLPM7T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T/action/storage_attestation","attest_author":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T/action/author_attestation","sign_citation":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T/action/citation_signature","submit_replication":"https://pith.science/pith/WP3SBP3E7JGJRLX474SXYLPM7T/action/replication_record"}},"created_at":"2026-07-05T10:59:13.225512+00:00","updated_at":"2026-07-05T10:59:13.225512+00:00"}