Developer Tools
symvanta.com
Index GitHub repositories into a live code knowledge graph, providing AI agents with exact symbols, call graphs, and blast radius over MCP.
ENDPOINT 1
https://mcp.symvanta.com/mcp
MCP server metadata
- Name
- symvanta-v1
- Version
- 1.0.0
Symvanta gives your AI coding agent your codebase's real call graph over MCP, so it stops guessing and knows what breaks before it edits. Do NOT shell out to grep, rg, find, cat, or ls on indexed repos. Use these tools (one call per question): behavior / 'how does X work' -> ask_codebase (scope:"all" cross-repo) symbol definition + signature -> find_node who calls X -> relate (kind:callers) what X depends on -> relate (kind:dependencies) what breaks if X changes -> relate (kind:blast_radius) what implements interface X -> relate (kind:implementers) type hierarchy of X -> relate (kind:heritage) trace polymorphism / DI, or class runtime order -> relate (kind:chain[, staysWithinClass:true]) how are X and Y connected -> relate (kind:path, selectors:[from, to]) literal identifier or string -> locate (mode:text; queries:[...] for several terms) HTTP route by path+method -> find_http_route env var / config key usage -> locate (mode:config; queries:[...] for several keys) repo or subtree skeleton -> map symbols in a file -> list_file_symbols what a diff / branch breaks -> diff_impact record/read the WHY (decisions) -> adr (op:record|list); find_node attaches them pattern discovery -> locate (mode:semantic) locate (mode:text) empty -> locate with no mode (auto-routes to semantic); never retry mode:text reworded Feedback: every ask_codebase answer carries an answer_id; report the outcome with feedback:{answerId,outcome:"useful"|"dead_end"|"corrected"} on any later tool call, and read the `lesson`/`lessons` hints on find_node / context / citations for known-good vs contested anchors. Source: the index returns filePath + line bounds + signature, not full source. Local Read only if the signature is insufficient; trust the live file over the index. Paths: every filePath is repo-relative logical (e.g. "src/user/email.js"), NOT "<repo>/<path>". Map it to your local layout before reading; pass the logical form for filePath inputs too. Branches: reads default to the default branch. Pin a feature branch with ref(op:"use", repository, branch), revert with ref(op:"clear"); the branch must be tracked first (same-repo PR, dashboard, or SessionStart hook). For UNCOMMITTED edits use ref(op:"index_working_tree", repository, changedFiles:[{path,content}]) to overlay + auto-pin (source and locate mode:semantic do NOT reflect it). commitSha overrides any pin. First call: init(repository:"owner/name") with your checkout's GitHub remote binds the session to the project holding it (calls without projectId then resolve there; ref op:"clear_project" unbinds). workspace.attached=false means this checkout is not indexed: attach it with add_repository (private repos: installation_id from list_installations), create_project first if it needs its own project, or work without the graph. Do not route an unattached checkout through the projects init lists.
Known tools 27
sourceRaw file/git access at the indexed commit, for cloud agents with no clone; with a clone prefer your own Read/Grep/git.
Inferred read-onlyadrArchitecture decision records: the WHY behind code, anchored to a symbol/file/project.
Inferred read-onlyestimate_scopePre-flight estimate before graph traversal: coarse count of files a task likely touches and whether it spans layers (common/browser/node/electron-main/.
Inferred read-onlymapToken-efficient repo (or subtree) skeleton: directory tree with each file's classes (methods in parens), functions, and HTTP endpoints, from the graph.
Inferred read-onlylist_tests_forTest suites and cases whose suite name matches a symbol name (case-insensitive substring).
Inferred read-onlyquick_lookupFast substring lookup of symbol DEFINITIONS by name: rows {name, kind, filePath, startLine, repositoryName} (displayName only when distinct from name), no ranking, no semantic.
Inferred read-onlyfreshnessReturns lastIndexedSha, lastIndexedAt, and the current remote HEAD sha for a repository: detect index drift from remote.
Inferred read-onlyfind_http_routeResolve an HTTP route by path + optional method to endpoint + handler (filePath + line bounds).
Inferred read-onlyask_codebaseSynthesis Q&A for behavior/architecture questions: answer + citations in one call.
Inferred read-onlyadd_repositoryAttach a GitHub repository to a project (owner + repo_name; clone URL derived; projectId defaults to the bound / default project).
Inferred read-onlycreate_projectCreate a project in the workspace (name; slug derived and made unique).
Potential side effectslist_installationsGitHub App installations connected to this workspace and the repositories each can reach (private included).
Inferred read-onlyreindex_repositoryDispatch a reindex job for an already-attached repository (repositoryId from list_repositories / add_repository).
Inferred read-onlyCONNECT WITH APPROVAL
Client installation
Review this server and its permissions before adding it. Secret placeholders must be set locally.
Codex
~/.codex/config.toml
[mcp_servers.symvanta-v1]
url = "https://mcp.symvanta.com/mcp"
enabled = true
Claude Code
.mcp.json
{
"mcpServers": {
"symvanta-v1": {
"type": "http",
"url": "https://mcp.symvanta.com/mcp"
}
}
}
Claude Desktop
Settings → Connectors → Add custom connector
Name: symvanta-v1
Remote MCP URL: https://mcp.symvanta.com/mcp
Add this remote URL as a custom connector in Claude Desktop. Availability depends on the user plan and workspace policy.
Cursor
.cursor/mcp.json
{
"mcpServers": {
"symvanta-v1": {
"url": "https://mcp.symvanta.com/mcp"
}
}
}
Visual Studio Code
.vscode/mcp.json
Add to Visual Studio Code{
"servers": {
"symvanta-v1": {
"type": "http",
"url": "https://mcp.symvanta.com/mcp"
}
}
}
Generic MCP
Client-specific MCP configuration
{
"name": "symvanta-v1",
"transport": "streamable-http",
"url": "https://mcp.symvanta.com/mcp"
}
MCP Inspector
Run the official MCP Inspector locally and enter the indexed Streamable HTTP endpoint.
TRUST AND VERIFICATION EVIDENCE
Loading Trust v2 evidence…
Checking the associated registrable domain. The BuiltWith key remains server-side.
Evidence is source-attributed and does not guarantee that a third-party server is safe. Risk labels are conservative metadata heuristics.