fgis-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| fgis_sourcesB | Available catalogues: FSNB, FER, TER registry, methodologies/coefficients, prices, archives and limits. |
| fgis_browse_sourceA | Navigate named FGIS source tree. Use returned GUID/id as parent and level for next branch. registry uses section + page; TER sections are 6 and 7. source_params supports region_id/zone_id/period_id/authority_id, stage, group_id, materials, year, fsnb_type. Current-price stages: authorities, services, worker_prices, group_indices, building_indices, direct_cost_indices. |
| fgis_diagnoseA | Check FGIS API JSON via configured route; explain proxy vs full-tunnel VPN limitations. |
| fgis_catalogB | List price regions; zones with region parent_id; periods with zone parent_id. |
| fgis_search_normsA | Search public norm API by code/text. Returns cards for all returned source publications. |
| fgis_read_normB | Read exact bare norm code online, including work steps and original resource quantities. |
| fgis_read_documentA | Read a public document/technical part online as text, without a dataset or disk cache. Use document_refs from browse: source='normative' + GUID for FSNB/FER/methods/PIR; source='fssc'/'fsem' without GUID for those full documents. Offsets are text characters. Pass prior provenance.sha256 as expected_sha256 to guard against changed documents. A bounded 15-minute RAM cache avoids repeated full downloads. refresh forces a new request. Embedded images are not OCR-processed. Use table tool for cells and merged spans. |
| fgis_search_documentA | Find literal case-insensitive text within one public document, without a dataset. Whitespace matches across paragraph/cell boundaries. Returns excerpts, character offsets and next_offset for more matches. Use read_document at a returned offset to read the surrounding text. No semantic search, coefficient selection or inference. Check text_status before interpreting no matches. |
| fgis_document_outlineA | List document paragraphs, explicit HTML headings and tables with text offsets and previews. offset/next_block_offset paginate blocks, not text characters. This is source structure, not an inferred official table of contents. Table indices feed fgis_read_document_table. |
| fgis_read_document_tableA | Read source table rows/cells online, preserving header flags, rowspan and colspan. Get table_index from outline. Each row paginates physical source cells with cell_offset/cell_limit; merged cells are not expanded. Long cell previews are explicitly truncated: read their text offsets with read_document. Includes surrounding text; no inferred price or coefficient applicability. |
| fgis_start_downloadB | Start durable download. Explicit queries, collection prefixes (each scans 01..99), price_books [{zone_id,period_id}]. No automatic all-Russia/history download. sources from fgis_sources traverses actual catalogues and downloads documents with technical parts. ['all_public'] selects all adapters; include_archive adds archived legal trees; all_periods includes historical split forms (large). max_tasks bounds traversal; bounded jobs retain pending tasks. TER registry is not TER table content; bulk archive files require interactive portal CAPTCHA. Search enumeration cannot establish exhaustive FSNB coverage. Returns job/dataset ID. |
| fgis_job_statusA | Read progress, failed tasks and interruption state of a background download. |
| fgis_cancel_jobA | Request cooperative cancellation after the current request/import; preserve downloaded data. |
| fgis_resume_jobB | Resume a stopped/partial job, verify saved source hashes, retry uncommitted tasks. |
| fgis_read_dataset_documentB | Read saved complete source JSON/technical parts in bounded character pages; no network. |
| fgis_list_datasetsA | List locally stored datasets without contacting FGIS. |
| fgis_dataset_infoB | Read dataset counts, coverage, manifest and local artifact paths. |
| fgis_query_datasetB | Offline norms/prices/documents search. Documents return summaries; read full content with document tool. |
| fgis_export_datasetA | Export stopped dataset to jsonl and/or parquet. SQLite always exists; return files and hashes. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| help_resource |
TDQS
Scored across 19 tools
Most tools target distinct resources and actions: source/catalog browsing, norm search vs read, online vs dataset reads, and job lifecycle tools are clearly separated. Some potential confusion exists between fgis_sources, fgis_catalog, and fgis_browse_source, and between offline query_dataset and online search_norms/search_document, but the descriptions are specific enough to disambiguate.
All tools share the fgis_ prefix and most follow a verb_noun pattern such as search_norms, read_document, cancel_job, and export_dataset. A few noun-style names like fgis_sources, fgis_catalog, fgis_dataset_info, and fgis_job_status, plus verb-only fgis_diagnose, break the pattern slightly, but the convention is still readable and predictable.
19 tools is on the heavy side and sits in the borderline range for a single MCP server. The tools are largely distinct, but catalog/source browsing, document reading, and job management could potentially be consolidated without losing much clarity.
The surface covers the full read-oriented workflow: discover sources and catalogues, browse and search online, start and manage downloads, then query, read, and export saved datasets. Minor gaps such as no direct way to fetch a dataset by ID from a job result and no dataset deletion/pruning keep it from a perfect score, but core workflows have no dead ends.