Connector guide
Use dVeracity in Claude
The dVeracity connector gives Claude a second opinion on sustainability data: it reads the canonical Open Footprint model, checks emissions and footprint payloads against the standard or regulation they claim to follow, and queries verified oil-and-gas emissions data. Every answer states which checks ran and which did not, so it can be handed to a reviewer instead of pasted into a chat.
This page documents the connector submitted to the Claude Connectors Directory. Tool names and costs were checked against the live server (0.8.3) on 2 November 2026.
What the connector does
dVeracity is a verification engine for regulated sustainability data. The connector exposes it as MCP tools, most of them read-only, grouped into four jobs:
- Design-time model reading. Entities, sectors, semantic codes and guardrails from the Open Footprint standard, so Claude can map an ERP or supplier feed onto a canonical model.
- Pre-submission validation. A payload is checked against the named standard or a sector's guardrails, and the response separates the checks that ran from the checks that could not.
- Regulatory evaluation. Approved EU CBAM policy rules are applied to import records and annual declarations, with the policy's provenance attached. See what the CBAM checks cover.
- Verified data queries. Natural-language questions over verified emissions and methane datasets (PPDM, OGMP 2.0).
The engine is domain-neutral; Open Footprint and EU CBAM are the standards onboarded today. It is not a carbon calculator, an emission-factor database, or a general climate assistant, and the tool descriptions tell Claude not to use it for those.
Connect Claude
The server is a remote streamable-HTTP MCP endpoint at https://api.dveracity.com/mcp. The same URL serves every user. There is nothing to install.
Claude.ai and Claude Desktop
- Open Settings → Connectors → Add custom connector.
- Paste the server URL
https://api.dveracity.com/mcp. - Leave client ID and secret blank. Claude identifies itself automatically, with a Client ID Metadata Document or dynamic client registration, and dVeracity's consent screen opens in your browser.
- Sign in to your dVeracity account and approve the scopes you want Claude to use.
Claude Code
claude mcp add --transport http dveracity https://api.dveracity.com/mcp
# then, inside Claude Code:
/mcpAny other MCP client that supports streamable HTTP can use the same URL. The public server card at https://api.dveracity.com/.well-known/mcp/server-card.json lists the transport, capabilities and the full tool catalogue before you connect.
Authentication and scopes
Discovery is public: initialize and tools/list answer without a credential so a client can see what dVeracity offers. Every tool call that touches an account needs an OAuth 2.1 bearer token or an X-Api-Key. The free Open Footprint tools are free of charge but are not anonymous.
The authorization server uses PKCE with S256 only and public clients only. Access tokens live one hour, refresh tokens thirty days, and authorization codes ten minutes. A token carries only the scopes approved on the consent screen:
| Scope | Grants |
|---|---|
semantic:templates | Question templates for the verified emissions data. |
semantic:query | Metered natural-language queries (1 credit per call). |
ofp:read | The Open Footprint model, sectors, policies and mapping tools. |
ofp:validate | Payload validation against Open Footprint or a standard (1 credit per call). |
compliance:read | Read approved regulatory policies and stored evaluations (paid plans). |
compliance:evaluate | Run a regulatory evaluation (paid plans). |
offline_access | Stay connected after this session ends without signing in again. |
Ask only for what the task needs. A token missing a scope is refused with HTTP 403 and insufficient_scope; reconnect and approve the scope to retry.
Tools and cost
Twenty-five tools. None changes or deletes records you already own: metered tools spend credits, compliance_evaluate stores a new evaluation record, and the billing tools open a checkout or subscription for a human to complete. Every tool advertises its price in the machine-readable tools/list response, which is public; the Semantic API manifest documents the equivalent REST endpoints.
Open Footprint model and catalogue — free
| Tool | What it does | Cost |
|---|---|---|
ofp_models | List the Open Footprint model domains. Start here for model questions. | Free |
ofp_search_entities | Find an Open Footprint entity such as a facility or product footprint. | Free |
ofp_entity | Read one entity's fields, types, keys and relationships. | Free |
ofp_sectors | List the industry sectors the model covers. | Free |
ofp_sector | Read one sector's classification, status and reference artifacts. | Free |
ofp_policies | Read the validation guardrails that apply to a sector. | Free |
ofp_semantics | List the semantic codes used to map source-system fields. | Free |
ofp_semantic_code | Find which model fields carry a given semantic code. | Free |
ofp_model_provenance | Report the Open Footprint version and commit being served. | Free |
semantic_templates | Show the question templates the verified data can answer. | Free |
list_standards | List the standards and versions validate_data can check. | Free |
keri_identity | Report the vLEI legal entity and agent identity attributed to a call. | Free |
credits_balance | Legacy balance lookup; use billing_status instead (removal planned for 0.9). | Free |
Validation and queries — metered
| Tool | What it does | Cost |
|---|---|---|
ofp_validate | Check an emissions or footprint record against Open Footprint or a sector's guardrails. | 1 credit ($0.01) |
validate_data | Check a payload against another supported standard. Call list_standards first. | 1 credit ($0.01) |
semantic_query | Ask a natural-language question over verified emissions data (PPDM, OGMP 2.0 methane). | 1 credit ($0.01) |
Regulatory checks — included with a VaaS plan
| Tool | What it does | Cost |
|---|---|---|
compliance_policies | List the approved regulatory policies available for evaluation. | VaaS plan |
compliance_policy | Read one policy's full rules, provenance and coverage. | VaaS plan |
compliance_policy_register | See which provisions of a regulation an evaluation checks, and which it does not. | VaaS plan |
compliance_evaluation | Retrieve or audit a previous regulatory evaluation by its id. | VaaS plan |
compliance_evaluate | Check an import record or an annual EU CBAM declaration against an approved policy. | VaaS plan |
Billing and plans
| Tool | What it does | Cost |
|---|---|---|
billing_plans | List the plans, prices and free allowance. This tool is anonymous. | Anonymous |
billing_status | Show the account balance, plan and free allowance. | Free |
billing_purchase | Start a credit purchase; returns a link the human must complete. | No charge |
billing_subscribe | Start a subscription checkout; the human completes it. | No charge |
When an account runs out of credits the server answers HTTP 402 with a machine-readable purchase pointer. The connector tells Claude to relay it to you; it does not retry.
Troubleshooting
401 — authentication required
Claude has no valid token. Approve the consent screen, or reconnect the connector so it can register and authorize again. Free tools still need a credential; only discovery (the tool list and billing_plans) is anonymous.
402 — payment required
The account is out of credits. Call billing_status to see what is missing, then let the human complete the checkout returned by billing_purchase or billing_subscribe. Do not retry the metered call until the balance is topped up.
403 — insufficient_scope
The token was approved without the scope that tool needs. Reconnect and approve the scope shown in the error; a different token cannot widen an existing grant on its own.
The connector cannot be added, or fails after consent
Check the URL is exactly https://api.dveracity.com/mcp with no trailing path. Remove the connector and add it again so a fresh client registration is issued; stale registrations from earlier attempts can fail even though the server is healthy. If it still fails, send us the time, the client and the error text — the OAuth error code is in our logs.
A tool returns “not configured” or a refusal
keri_identity answers “not configured” when no vLEI identity is set up; that is a valid response, not an outage. A scope refusal and a 402 are also deliberate answers with a documented next step.
Timeouts
The endpoint is hosted in Google Cloud. Validation and natural-language queries can take a few seconds; model and catalogue lookups are fast. Retry once on a network error, then check the Trust page for the current status of our controls.
Data handling
The connector never changes or deletes data you already have. Three tools write: compliance_evaluate stores a new evaluation record (your audit trail), and billing_purchase and billing_subscribe create a checkout or subscription session that a person must complete. All other tools only read. Metered calls are attributed to the account or key that paid for them, together with the client's declared name and version. Claude receives the tool results; what it does with them is governed by Anthropic's policies. Our own handling is described in the Privacy Policy and the Trust & Security page.
Support and source
- Support — how to reach us, and what to include in a bug report.
- Semantic API overview — the wider API, pricing and vLEI identity.
- Machine-readable manifest — auth, per-call costs and HTTP 402 semantics.
- MCP server source — the open-source package behind the tools.