Cost Guard MCP
MCP serverby mcpsmithsSRC
Pre-flight query cost and result-size guardrails for AI agents on BigQuery, Snowflake, Databricks
SRC Publisher-stated: these are the publisher's own words, not an assessment.
Key facts
- Type
- MCP server · mcp-localSRC
- Publisher
- mcpsmiths · organizationSRC
- Licence
- MITSRC
- Open source
- YesDER
- Hosting
- Local (runs on your machine)SRC
- Latest version
- 0.6.0SRC
- Languages
- pythonSRC
- First published
- SRC
- In this catalogue since
- DER
- Last verified
- Not verified by a researcher. Last seen in a source on 2026-11-21.
Capabilities and categories
Assigned by keyword rules from the description unless marked otherwise. The matched word is shown so you can judge the match.
| Capability | Evidence | Kind |
|---|---|---|
| Agent Governance | matched guardrails | DER |
| Data Warehouses | matched bigquery | DER |
| Cloud Cost Optimization | matched cost optimization | DER |
| Policy Enforcement | matched guardrails | DER |
| Warehouse Querying | matched bigquery | DER |
Filed underAgent InfrastructureCloud & Platform IntegrationsData & KnowledgeAgent SecurityCloud ProvidersData Engineering & AnalyticsCloud Infrastructure
Integrations DER
Platform compatibility
No pairing has been tested by AgentSidekick. Each status says what kind of evidence exists, not whether it works for you.
Protocol-level29
Each platform listed officially documents support for MCP. This specific pairing has not been tested.
Platform discontinued1
| Platform | Basis |
|---|---|
| Roo CodeDiscontinued | Roo Code documented support for MCP but is discontinued. Project shut down and repository archived. |
Unknown17
| Platform | Basis |
|---|---|
| OpenClaw | Could not confirm whether OpenClaw supports MCP. |
| OpenHands | OpenHands documents MCP support, but it runs only on web, api, slack surfaces and this server runs locally as a process on your machine. Whether OpenHands can reach a local server is not established. |
| Aider | Could not confirm whether Aider supports MCP. |
| Bolt (bolt.new) | Bolt (bolt.new) documents MCP support, but it runs only on web surfaces and this server runs locally as a process on your machine. Whether Bolt (bolt.new) can reach a local server is not established. |
| Amazon Q DeveloperWinding down | Could not confirm whether Amazon Q Developer supports MCP. |
| Amazon Bedrock Agents (Agents Classic)Maintenance only | Could not confirm whether Amazon Bedrock Agents (Agents Classic) supports MCP. |
| GPTs (custom GPTs)Winding down | Could not confirm whether GPTs (custom GPTs) supports MCP. |
| Gemini (app) | Gemini (app) documents MCP support, but it runs only on web, mobile surfaces and this server runs locally as a process on your machine. Whether Gemini (app) can reach a local server is not established. |
| Gemini Enterprise Agent Platform (formerly Vertex AI / Vertex AI Agent Builder) | Gemini Enterprise Agent Platform (formerly Vertex AI / Vertex AI Agent Builder) documents MCP support, but it runs only on web, api surfaces and this server runs locally as a process on your machine. Whether Gemini Enterprise Agent Platform (formerly Vertex AI / Vertex AI Agent Builder) can reach a local server is not established. |
| GitHub Copilot ExtensionsDiscontinued | Could not confirm whether GitHub Copilot Extensions supports MCP. |
| Google AI Studio | Could not confirm whether Google AI Studio supports MCP. |
| JulesWinding down | Could not confirm whether Jules supports MCP. |
| Microsoft Copilot | Could not confirm whether Microsoft Copilot supports MCP. |
| Microsoft Copilot Studio | Microsoft Copilot Studio documents MCP support, but it runs only on web surfaces and this server runs locally as a process on your machine. Whether Microsoft Copilot Studio can reach a local server is not established. |
| Replit Agent | Replit Agent documents MCP support, but it runs only on web surfaces and this server runs locally as a process on your machine. Whether Replit Agent can reach a local server is not established. |
| Sourcegraph Cody | Could not confirm whether Sourcegraph Cody supports MCP. |
| v0 | v0 documents MCP support, but it runs only on web, api surfaces and this server runs locally as a process on your machine. Whether v0 can reach a local server is not established. |
Install
Read from registry metadata. Follow the publisher's instructions; we have not run this.
| Registry | Package | Transport | Runtime |
|---|---|---|---|
| pypi | cost-guard-mcp | stdio | not stated |
Environment variables SRC
| Name | Required | Secret | Publisher's description |
|---|---|---|---|
| GOOGLE_APPLICATION_CREDENTIALS | No | Secret | BigQuery: path to a service-account key file (Application Default Credentials). Required to use the bigquery engine. |
| SNOWFLAKE_ACCOUNT | No | No | Snowflake: account identifier. Required to use the snowflake engine. |
| SNOWFLAKE_USER | No | No | Snowflake: username. Required to use the snowflake engine. |
| SNOWFLAKE_ROLE | No | No | Snowflake: role to assume. Required to use the snowflake engine; has no default and must never be ACCOUNTADMIN (ACCOUNTADMIN, SECURITYADMIN and ORGADMIN are rejected before any Snowflake connection is opened - the check runs when the Snowflake configuration is first loaded, on the first Snowflake tool call; the server process itself starts regardless, so a bad role surfaces as a failed check_credentials or a tool error, not as a startup exit). |
| SNOWFLAKE_SECONDARY_ROLES | No | No | Snowflake: secondary roles activated at login. NONE (default) means only SNOWFLAKE_ROLE is active in the session; ALL is the explicit opt-out that activates every role granted to SNOWFLAKE_USER. |
| SNOWFLAKE_PRIVATE_KEY_PATH | No | Secret | Snowflake: path to an RSA private key file for key-pair auth (preferred over SNOWFLAKE_PASSWORD). Either this or SNOWFLAKE_PASSWORD is required to use the snowflake engine. |
| SNOWFLAKE_PRIVATE_KEY_PASSPHRASE | No | Secret | Snowflake: passphrase for an encrypted SNOWFLAKE_PRIVATE_KEY_PATH key file, if the key is encrypted. |
| SNOWFLAKE_PASSWORD | No | Secret | Snowflake: password (discouraged; prefer SNOWFLAKE_PRIVATE_KEY_PATH). Either this or SNOWFLAKE_PRIVATE_KEY_PATH is required to use the snowflake engine. |
| SNOWFLAKE_EDITION | No | No | Snowflake: account edition that sets the USD/credit price of estimates - standard (default), enterprise, business_critical or vps. The per-call edition argument can only raise it. |
| DATABRICKS_SERVER_HOSTNAME | No | No | Databricks: SQL warehouse server hostname (e.g. my-workspace.cloud.databricks.com, no scheme, no trailing slash). Required to use the databricks engine. |
| DATABRICKS_HTTP_PATH | No | No | Databricks: SQL warehouse HTTP path (from the warehouse's Connection Details tab, e.g. /sql/1.0/warehouses/<warehouse-id>). Required to use the databricks engine. |
| DATABRICKS_TOKEN | No | Secret | Databricks: personal access token (simplest auth option). Either this or DATABRICKS_CLIENT_ID + DATABRICKS_CLIENT_SECRET is required to use the databricks engine. |
| DATABRICKS_CLIENT_ID | No | No | Databricks: OAuth machine-to-machine service-principal client ID (preferred for automated use). Requires DATABRICKS_CLIENT_SECRET as well. |
| DATABRICKS_CLIENT_SECRET | No | Secret | Databricks: OAuth machine-to-machine service-principal client secret. Requires DATABRICKS_CLIENT_ID as well. |
| COST_GUARD_ALLOW_MULTI_STATEMENT_SQL | No | No | Set to 1 to let estimate_query_cost/run_query_bounded accept text containing more than one SQL statement. Off by default: the cost cap is only known to cover the single statement the dry run/EXPLAIN priced. Skips only this server's scanner: Snowflake still rejects multi-statement text (sessions are pinned to MULTI_STATEMENT_COUNT=1), and only a single SELECT/WITH/VALUES statement is wrapped in a LIMIT subquery - a script runs unwrapped and its row cap (max_rows, default 1000) is enforced by the paged fetch. |
| COST_GUARD_ALLOW_WRITE_STATEMENTS | No | No | Set to 1 (exactly that value) to let estimate_query_cost/run_query_bounded accept statements that are not reads. Off by default: only text whose first keyword (after comments and opening parentheses) is SELECT, WITH, VALUES, SHOW, DESCRIBE, DESC or EXPLAIN is estimated or run - a WITH must be followed by a SELECT or VALUES - because warehouse rows re-enter the model's context and can steer the next call. The gate is defence in depth: the authorisation boundary is the credential's grants (deploy a read-only principal). With COST_GUARD_ALLOW_MULTI_STATEMENT_SQL=1 every statement in the text must be a read, so BigQuery scripts (DECLARE/SET/BEGIN) need both opt-outs. Any other value leaves the gate on. |
| COST_GUARD_MAX_IN_FLIGHT_QUERIES | No | No | Maximum number of run_query_bounded calls allowed in flight at once (positive integer, default 8). A cancelled call keeps its slot until its warehouse query completes or the 120 s watchdog fires; a call over the cap fails with a tool error naming this variable before any warehouse connection is opened. |
| COST_GUARD_DATABRICKS_ALLOW_WAREHOUSE_START | No | No | Set to true (exactly that value) to let the read-only Databricks tools (check_credentials, estimate_query_cost) open a session on a SQL warehouse that is not RUNNING, which starts the warehouse and bills compute until it auto-stops. Off by default: the server first reads the warehouse's state through the Warehouses API (GET /api/2.0/sql/warehouses/<id>, CAN_VIEW/CAN_USE suffices) and, unless it is RUNNING - also when the state cannot be read - check_credentials returns ok=false and estimate_query_cost returns no byte or dollar figure, without opening a session. run_query_bounded is not gated: its own pre-execution estimate and the execution may start the warehouse. Any other value leaves the guard on. |
| COST_GUARD_SNOWFLAKE_ALLOW_WAREHOUSE_RESUME | No | No | Set to true (exactly that value) to let the read-only Snowflake tools run SELECT CURRENT_REGION(), the INFORMATION_SCHEMA.QUERY_HISTORY read and check_credentials' session-context SELECT on a warehouse that is not STARTED, which can resume a suspended AUTO_RESUME warehouse and bill compute until it auto-suspends. Off by default: those statements are skipped unless the pinned warehouse's SHOW WAREHOUSES state is STARTED (also skipped when no warehouse is pinned or the state cannot be read), and the QUERY_HISTORY read is skipped whenever estimate_query_cost is called without a warehouse argument; estimate_query_cost then prices at Gen1 rates with the byte-tier heuristic and says so in a caveat, and check_credentials returns ok=true with a detail built from the SHOW WAREHOUSES row. USE WAREHOUSE, EXPLAIN USING JSON, SHOW WAREHOUSES and SELECT CURRENT_WAREHOUSE() always run; run_query_bounded is not gated. Any other value leaves the guard on. |
Permissions and signals
Signals, not a safety verdict. They describe what the product's metadata references. Security review: not reviewed.
| Signal | Evidence | How observed | Observed |
|---|---|---|---|
| Runs a local process | runs locally via pypi | registry metadata | 2026-11-21 |
| Asks for a secret | declares secret env var(s): GOOGLE_APPLICATION_CREDENTIALS, SNOWFLAKE_PRIVATE_KEY_PATH, SNOWFLAKE_PRIVATE_KEY_PASSPHRASE, SNOWFLAKE_PASSWORD, DATABRICKS_TOKEN | registry metadata | 2026-11-21 |
Known advisories for the published package SRC
An advisory existing does not mean the listed version is affected: compare the vulnerable range with the version you install. None found is not evidence of safety.
No advisories found
No published advisory matches a package this product declares. Products without a declared package cannot be matched at all. This is not evidence of safety.
Evaluation
Separate dimensions, each with its evidence. There is no overall score: the dimensions do not add up to one number honestly.
- Source authority
Publisher-submitted
Published to the official MCP Registry by the namespace owner. Field values are the publisher's own claims.
- Publisher verification
Namespace verified by registry
The MCP Registry verifies namespace ownership (GitHub login or DNS) before publishing. This confirms who published it, not what it does.
- Maintenance activity
Active
Last push 2026-11-18 (3 days before this page was built).
- Licence clarity
MIT
Licence identifier as reported by GitHub repository metadata.
- Installation clarity
Machine-readable
1 package(s) and 0 remote endpoint(s) declared in registry metadata.
- Documentation
Repository / website only
No dedicated documentation URL on record. AgentSidekick has not assessed documentation quality.
- Compatibility evidence
Protocol-level only
Compatibility is inferred from the standard it implements. No pairing has been tested.
- Security review
Not reviewed
AgentSidekick has not audited this product. 2 automated signal(s) are listed under Permissions & signals — they describe what it references, not whether it is safe.
Pricing
Not published in any source we track. The code is open source, which says nothing about the cost of any hosted service.
Repository activity
mcpsmiths/cost-guard-mcp. Read from the GitHub API on 2026-11-21.
Version history SRC
- 0.6.0
- 0.5.0
- 0.3.2
- 0.3.0
- 0.2.0
- 0.1.1
- 0.1.0
Change history
What changed between two ingests of the same source, and when the product was first seen. Registry releases are under Version history.
- First seen
Competitors and alternativesDER
Paired by shared capabilities and integrations. Similar on paper; not tested side by side.
Direct
Same type, overlapping capabilities.
MCP serverby camber-mcp.cambercloud.comSRC
SharesData WarehousesWarehouse QueryingBigQueryDatabricks
MCP serverby aliengiraffe.aiSRC
SharesAgent GovernancePolicy EnforcementData WarehousesWarehouse Querying
MCP serverby Optisol-BusinessSRC
SharesData WarehousesWarehouse QueryingBigQuerySnowflake
MCP serverby kalehdooSRC
SharesData WarehousesWarehouse QueryingBigQuerySnowflake
MCP serverby SnowLeopard-AISRC
SharesData WarehousesWarehouse QueryingBigQuery
MCP serverby Snowflake-LabsSRC
SharesData WarehousesWarehouse QueryingSnowflake
Adjacent
A different form or a partial overlap.
Pluginby altimateaiSRC
SharesData WarehousesWarehouse QueryingCloud Cost OptimizationBigQuery
Pluginby gemini-cli-extensionsSRC
SharesData WarehousesWarehouse QueryingBigQuery
Pluginby GRCEngClubSRC
SharesData WarehousesWarehouse QueryingSnowflake
Provenance
Which source set each field, what kind of claim it is, and when.
| Field | Source | Kind | Set on |
|---|---|---|---|
| Canonical URL | Official MCP Registry | SRC | 2026-11-21 |
| Publisher | Official MCP Registry | SRC | 2026-11-21 |
| Description | Official MCP Registry | SRC | 2026-11-21 |
| Hosting | Official MCP Registry | SRC | 2026-11-21 |
| Integration type | Official MCP Registry | SRC | 2026-11-21 |
| Languages | GitHub repository metadata | SRC | 2026-11-21 |
| Latest version | Official MCP Registry | SRC | 2026-11-21 |
| Licence | GitHub repository metadata | SRC | 2026-11-21 |
| Name | Official MCP Registry | SRC | 2026-11-21 |
| Open source | GitHub repository metadata | DER | 2026-11-21 |
| Type | Official MCP Registry | SRC | 2026-11-21 |
| Published | Official MCP Registry | SRC | 2026-11-21 |
| Repository | Official MCP Registry | SRC | 2026-11-21 |
| source_updated_at | Official MCP Registry | SRC | 2026-11-21 |
| Source URL | Official MCP Registry | SRC | 2026-11-21 |
Sources
- Official MCP Registry: registry.modelcontextprotocol.io (external)checked
Not verified by a researcher. Last seen in a source on . Report incorrect information