Argosvix MCP server lets an agent query and act on your LLM observability data โ costs, alerts, expensive calls, anomalies โ directly from a conversation, across 89 tools.
Copy the AI prompt to install this server into Claude Code, Cursor, or another agent โ or use 1-click editor setup below.
This server is confirmed live โ we successfully called its tools/list endpoint directly (see the verified badge above). We haven't yet sandbox-tested the stdio install command below specifically, which is a separate, ongoing check.
๐ก Paste the JSON block into your client's configuration file under mcpServers, then restart the application.
Inspect callable tools, capabilities, and parameters exposed to AI agents by MCP Server.
query_callsRetrieve recent LLM call records captured by Argosvix. Filterable by provider / model / time range / tag (tagKey + tagValue pair). Defaults to the last 24 hours, 100 records.
get_cost_summaryReturn cost / call count / token aggregates per time range, with a per-provider breakdown. When groupBy="none" is specified, a per-provider breakdown is still returned for backend compatibility (check the response.total field for the overall sum).
list_alertsReturn the list of configured alerts plus trigger history within the last 24 hours.
list_proposalsReturn the unresolved improvement proposals found by the Argosvix guardian (quality drift / reliability anomalies / cost switching / safety / silencing noisy alerts). Approving, dismissing, and executing happen in the dashboard inbox (agents can only read and converse).
get_proposal_threadReturn the thread for a proposal (the questions asked so far and the AI's replies). Get proposalId from list_proposals.
reply_proposalPost a question about a proposal and get the AI's reply (same as the inbox conversation). Explanation only โ nothing is executed. Get proposalId from list_proposals.
Argosvix MCP server lets AI agents (Claude Desktop, Cursor, Codex CLI, custom MCP clients) query, manage, and operate their LLM observability data directly from the conversation. Supports both stdio (subprocess) and HTTP (remote / self-host) transports.
Surface: 89 tools (86 generally available + 3 internal operations tools that return 403 for customer accounts) / 3 resources / 8 resource templates / 3 prompts. Health and anomaly endpoints (get_account_health / detect_anomaly / propose_alert_rules / classify_calls_batch / propose_eval_criteria) plus a runtime control plane (budget gates / policy gates / human-approval gates) let an agent both observe and act. Release history is available on npm.
You're already sending LLM calls through @argosvix/sdk or the Python SDK. Now ask Claude / Cursor questions like:
No dashboard tab-switching. The agent fetches the data via this MCP server using your Argosvix API key.
One-click (starts keyless; set ARGOSVIX_API_KEY afterwards to use the tools for your own account):
Claude Code users can install the argosvix plugin instead โ it bundles this server plus setup skills:
Manual:
Edit your Claude Desktop config:
~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.jsonRestart Claude Desktop. All 89 tools appear under the argosvix__ prefix (e.g. argosvix__query_calls, argosvix__get_account_health, argosvix__create_budget_gate).
No API key yet? The server also starts without
ARGOSVIX_API_KEYin introspection-only mode: all 89 tools are listed so you can evaluate the surface, and two public data tools (get_synth_daily/get_daily_readthroughโ Argosvix's own daily AI digest and OSS readthrough) work with no key at all. Every other tool call returns instructions for getting a key at https://dashboard.argosvix.com/api-keys.Notes on the public data tools: they are part of the default
fullprofile only (not inARGOSVIX_MCP_PROFILE=core), and like every other tool they honorARGOSVIX_API_BASEโ if you point the server at a different base, they fetch/v1/synth-daily//v1/readthroughfrom that base without authentication.
Edit ~/.cursor/mcp.json:
ARGOSVIX_MCP_PROFILE)The default profile is full (all 89 tools). If your MCP client's context budget is tight, set ARGOSVIX_MCP_PROFILE=core to expose only the 11 essentials for day-to-day operations:
query_calls ยท aggregate_calls ยท get_cost_summary ยท get_percentiles ยท get_account_health ยท detect_anomaly ยท list_alerts ยท create_alert ยท silence_alert ยท unsilence_alert ยท get_deployed_prompt
Works in both stdio and HTTP transports. Unknown values fall back to full with a warning on stderr.
ARGOSVIX_MCP_LANG)Tool, resource, and prompt descriptions returned by tools/list / resources/list /
prompts/list are in English by default. Set ARGOSVIX_MCP_LANG=ja to get the original
Japanese descriptions instead. Unset or unknown values fall back to en (with a warning
on stderr for unknown values). Tool names, input schemas, and behavior are identical in
both languages โ only the human/agent-facing description text changes.
Works in both stdio and HTTP transports.
89 tools in total โ the full list is returned by tools/list. A sample of the core read / write surface:
| Tool | Purpose | Type |
|---|---|---|
query_calls | Recent LLM call records, filterable by provider / model / time range | read |
get_cost_summary | Aggregate cost / calls / tokens by provider or model | read |
list_alerts | Configured alerts + recent trigger status | read |
get_alert | Detail of a specific alert + recent trigger history | read |
list_alert_events | Alert trigger events across the account (notification history) | read |
silence_alert | Mute a specific alert (= temporary notification stop, default 24h) | write |
unsilence_alert | Resume notifications for a previously muted alert | write |
create_alert | Create a new alert rule (cost / error rate / latency / anomaly) | write |
acknowledge_alert | Mark a specific alert event as acknowledged (idempotent, orthogonal to silence) | write |
Resources expose read-only snapshots that AI agents can pull into context without an explicit tool call.
| URI | Purpose |
|---|---|
argosvix://account | Plan / quota / current-month record usage / retention snapshot (non-sensitive, Bearer-only) |
argosvix://alerts/active | Snapshot of currently enabled alerts |
argosvix://cost/today | Last-24h cost breakdown by provider (with response.total) |
Resource templates let agents construct dynamic URIs from a known id. The server enforces account scope via PK lookup on the backend, so cross-account ids return 404.
| URI template | Purpose |
|---|---|
argosvix://calls/{id} | Single LLM call record (provider / model / tokens / cost / latency / tags / error / trace_id). Plug in any id from query_calls results. |
argosvix://alerts/{id} | Single alert rule (name / type / threshold / window / channelKinds / sleep / enabled / silencedUntil) + recent 20 trigger events. Plug in any id from list_alerts results. channelTargets (notification destinations) is structurally dropped. |
argosvix://traces/{id} | Single trace = all spans grouped by trace_id (LLM call time-series). Top 50 spans only (LLM context budget cap); errorDetails / requestMeta are structurally dropped. Plug in any traceId from query_calls results. |
Prompts are reusable templates the user can launch as slash commands.
| Name | Purpose |
|---|---|
cost_review | Compare 24h / 7d / 30d cost trends and flag anomalies |
alert_audit | Audit current alert rules and propose improvements |
incident_triage | Investigate recent error / latency anomalies (default last 24h) |
resources.subscribe capability is declared in stdio mode. Clients can subscribe to one or more of the static resources below; the server polls each subscribed resource every 60 seconds and emits notifications/resources/updated when the content hash changes.
Subscribable URIs (resource templates such as argosvix://calls/{id} are not subscribable):
argosvix://accountargosvix://alerts/activeargosvix://cost/todayHTTP transport (= argosvix-mcp --http) does not declare subscribe and rejects subscribe requests, since per-request stateless mode cannot keep a subscription set or deliver server-initiated notifications. Use stdio for live updates.
listChanged is intentionally not declared: the resource list is fixed for the server lifetime.
Implementation notes (v0.10):
readResource calls).isShuttingDown flag, so notifications are not emitted for URIs removed mid-cycle.ARGOSVIX_MCP_DEBUG=1 the server logs { uri, errorClass } to stderr โ error messages are intentionally not logged to avoid leaking upstream payload text.McpError(InvalidParams) + axis 4 Tier 1 dispatcher coverage).The MCP server sends queries to https://ingest.argosvix.com using your API key. No prompts or completions are exposed โ only metadata (tokens, cost, latency, model name, your tags).
MIT ยฉ Yuto Makihara (Argosvix). See LICENSE.
Factual signals from GitHub, npm, and our automated checks โ not a rating.
No reviews yet โ be the first to share how this listing worked for you.
Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.
[](https://allmcps.com/mcp/argosvix-mcp-server)<a href="https://allmcps.com/mcp/argosvix-mcp-server"><img src="https://allmcps.com/api/badge/argosvix-mcp-server?style=directory" alt="MCP Server on AllMCPs" /></a>