Skip to content

mcp-tools.md

Latest commit

 

History

History
96 lines (84 loc) · 5.92 KB

mcp-tools.md

File metadata and controls

96 lines (84 loc) · 5.92 KB
Raw

🤖 MCP Tools Reference

SageFs exposes dozens of MCP tools — from send_fsharp_code to run_tests to discover_features. Any MCP client can connect.

Connection

Streamable HTTP (recommended — auto-reconnects, no session drops):

{ "mcpServers": { "sagefs": { "type": "streamable-http", "url": "http://localhost:37749/" } } }

SSE (legacy clients that don't support Streamable HTTP yet):

{ "mcpServers": { "sagefs": { "type": "sse", "url": "http://localhost:37749/sse" } } }

Full Tool List

Tool Description
send_fsharp_code Execute F# code. Each ;; is a transaction boundary.
check_fsharp_code Type-check a snippet without executing. Code is checked in the current FSI session context — prior send_fsharp_code definitions are in scope, but namespaces must be explicitly opened. "Not defined" errors usually mean a missing open, not a real bug.
get_completions Code completions at cursor position.
cancel_eval Cancel a running evaluation.
load_fsharp_script Load .fsx with partial progress.
get_recent_fsi_events Recent evals, errors, loads with timestamps.
get_fsi_status Session health, loaded projects, affordances.
get_startup_info Projects, features, CLI arguments.
get_available_projects Discover .fsproj/.sln/.slnx files.
explore_namespace Browse types in a .NET namespace.
explore_type Browse members of a .NET type.
get_elm_state Current UI render state.
reset_fsi_session Soft reset — clear definitions, keep DLLs.
hard_reset_fsi_session Full reset — rebuild, reload, fresh session.
create_session Create an isolated FSI session.
list_sessions List all active sessions.
stop_session Stop a session by ID.
switch_session Switch active session.
switch_workflow Switch between REPL (interactive) and Live (WebLive) workflows.
enable_live_testing Turn on live unit testing.
disable_live_testing Turn off live unit testing.
get_live_test_status Test state with optional file filter.
run_tests Run tests by pattern or category. Waits up to 15s for hot reload to complete first.
set_run_policy Auto-run policy per category (every/save/demand/disabled).
set_test_timeouts Configure per-test and global run timeouts.
get_test_trace Test cycle timing waterfall.
explain_test_run Why a test was selected to run — trigger reason, changed symbols, flaky status.
explain_test_failure Enriched failure context for a test that recently went Passed→Failed.
query_test_coverage Which tests transitively cover a given symbol via the dependency graph.
get_file_coverage Per-line coverage data for a file — bitmap + dependency graph synthesis.
visualize_domain_model Visualize a discriminated union type as a state machine diagram.
list_tests List all discovered tests, optionally filtered by pattern or file path. Returns grouped-by-file results with source locations.
get_cell_dependencies Expose the cell dependency graph with staleness annotations. Shows which cells are stale and why.
discover_features Context-aware feature discovery. Ranks available SageFs features by relevance to current session state.
Analysis & Diagnostics
diagnose Full diagnostic report combining test failures, cell staleness, ripple plan, suggestions, and performance.
coverage_intel Analyze test coverage quality — find blind spots, correlate failures, assess diagnostic power.
impact_forecast Forecast performance impact for cells — detect regressions, measure downstream blast radius.
suggest_next_action Prioritized "what should I do next?" queue combining coverage, impact, and staleness data.
suggest_repair Given a failing test, trace causal changes and suggest which symbol to fix.
suggest_next_cell Type-directed suggestions for what to evaluate next based on current bindings in scope.
plan_ripple Plan cascade re-evaluation for changed cells using the live dependency graph.
preview_what_if Preview what would change if a binding had a different value — without executing.
decompose_pipeline Decompose an F# pipeline into stages with purity classification (pure/effectful/unknown).
Export & Session History
export_notebook Export session as a notebook-style .fsx file with cell metadata.
export_session_transcript Export session as a clean, topologically-sorted .fsx transcript.
get_session_filmstrip Visual history of all evaluations — each as a "frame" with code, bindings, and duration.
get_eval_timeline Performance sparkline and percentile statistics (P50/P95/P99) for eval durations.
get_eval_diff Before/after diff comparison of recent evaluation outputs.
get_message_journal Structured audit log of eval events, filterable by severity and source.
manage_scratch_pad View, export, or promote ephemeral code snippets from the session history.

Per-Client Config Examples

GitHub Copilot (CLI) — ~/.copilot/github-copilot/mcp.json:

{ "servers": { "sagefs": { "type": "sse", "url": "http://localhost:37749/sse" } } }

Claude Code — ~/.claude/claude_desktop_config.json:

{ "mcpServers": { "sagefs": { "type": "sse", "url": "http://localhost:37749/sse" } } }

Windsurf / Cursor — .cursor/mcp.json or Windsurf MCP settings:

{ "mcpServers": { "sagefs": { "type": "sse", "url": "http://localhost:37749/sse" } } }

OpenCode — mcp.json:

{ "mcpServers": { "sagefs": { "url": "http://localhost:37749/sse" } } }

Works with GitHub Copilot (CLI & VS Code), Claude Code, Claude Desktop, OpenCode, Windsurf, Cursor, and any MCP-compatible tool. The edit → auto-test → poll workflow means agents don't even need to call eval — just edit files and check get_live_test_status.