MCP Server FAQs
General
1. What is the Model Context Protocol (MCP)?
The Model Context Protocol (MCP) is a machine-readable framework developed by Anthropic that allows large language models to query structured data in a consistent and governed way. It translates natural language prompts into standardized data requests, enabling models to access relevant insights without relying on unstructured retrieval or custom integrations. Read more here.
2. What is the Carbon Arc MCP Server?
The Carbon Arc MCP Server is an infrastructure layer that allows AI agents and applications to securely query Carbon Arc's structured data through MCP tool calls. It transforms natural language inputs into composable, auditable data queries that return machine-readable outputs for downstream analysis or decision-making.
3. How does natural language querying work with Carbon Arc data?
Carbon Arc's MCP Server converts natural language queries into a structured framework object, composed of elements such as entities, insights, time, space, and optional filters. These structured framework objects are used to query Carbon Arc data assets and return data and context to the user.
4. What types of data can be accessed through the Carbon Arc MCP Server?
Carbon Arc's catalog of data assets is available for access through the Carbon Arc MCP Server (where permitted by each data provider). Our Library can be accessed here.
5. What is Lenses?
Lenses is a Carbon Arc product native to the Carbon Arc web application that is powered by the Carbon Arc MCP Server. It allows users to ask natural language queries through persona-based "Lenses" (e.g., Investor, Consultant) that translate prompts into MCP tool calls to generate contextualized insights and decision-ready outputs based on Carbon Arc's data assets.
6. How does Lenses use the Carbon Arc MCP Server?
Lenses is the user interface that makes Carbon Arc's MCP Server accessible within the Carbon Arc platform. It uses the Carbon Arc MCP Server to query and respond to natural language prompts by orchestrating MCP tool calls.
7. What foundation models are used within Lenses?
Lenses uses foundation models provided through AWS Bedrock, including Anthropic's Claude models, to process and summarize structured data retrieved via the Carbon Arc MCP Server.
8. Is Lenses a standalone product or a feature within the Carbon Arc platform?
Lenses is a feature within the Carbon Arc platform and is native to the Carbon Arc web application. Lenses requires an active Carbon Arc subscription for "Professional" and "Business" plans. Enterprise plans require an active Carbon Arc subscription and a Carbon Arc MCP Server subscription.
9. What do you mean by "native" versus "non-native" access of Carbon Arc's server?
Carbon Arc's MCP server can be accessed directly within the Carbon Arc web application through Carbon Arc's product called "Lenses". Carbon Arc's MCP Server can also be accessed by setting up a custom connector (five step process) in ChatGPT, Claude, or other LLM's of choice.
Access & Usage
10. How can I access the Carbon Arc MCP Server?
You can access Carbon Arc's MCP Server in three ways:
-
Lenses: Carbon Arc's native user interface within the web application (no additional setup required)
-
Custom Connector using an LLM (e.g., ChatGPT, Claude): please find step-by-step directions on how to setup the Carbon Arc MCP Server with your Carbon Arc credentials here.
-
Programmatically: minimal, self-contained MCP client examples showing how to connect from every popular language using the official Model Context Protocol SDKs
11. How do I get started now that I have access?
Simply type a prompt with your use case and ask what prompts might be interesting: "I'm interested in trading WMT before earnings, show me what prompts I can ask to learn more about Walmart's performance?"
12. How do I know what tools are available in the Carbon Arc MCP Server?
Simply prompt the Carbon Arc MCP Server with "Show me the tools available in the Carbon Arc MCP Server" or find more information here.
13. What is the difference between the CarbonArcMCP and CarbonArcResearchMCP endpoints?
CarbonArcMCP provides our core toolset. CarbonArcResearchMCP provides MCP research tools designed to support the full demand lifecycle. Entity Research provides structured overviews of companies, brands, and retailers with competitive context and coverage. Forecast evaluates forward-looking dynamics ahead of earnings. Acquire, Retain, Target, and Upsell tools analyze customer acquisition, loyalty and churn, segmentation, and expansion opportunities helping users move from raw data to actionable insights across the growth funnel.
Both are available in Lenses and non-natively and are included in the subscription.
14. What is "research mode" in Lenses?
Research mode evaluates a larger range of data assets and generates comprehensive reports. It is optimal for extensive analysis on specific entities, companies or categories. Research mode may consume more Carbon Arc MCP tokens than regular tool calls.
15. What is the difference between accessing Carbon Arc's MCP Server using a custom connector setup through an LLM (e.g., Claude or ChatGPT) and Lenses?
A custom connector in Claude or ChatGPT uses the same Carbon Arc MCP Server as Lenses. Features & capabilities and data privacy, handling, & compliance may be different and details can be found below.
16. How much does access to the Carbon Arc MCP Server cost?
The Carbon Arc MCP Server is included in the monthly or annual "Professional" and "Business" subscriptions, where pricing information can be found here. Enterprise plans have custom pricing tailored to an organizations' needs.
17. Do I pay extra for Lenses or to access the Carbon Arc MCP Server through a custom connector?
No, there is no additional fee beyond the "Professional", "Business" or custom Enterprise monthly / annual subscriptions.
18. Are there any seat limitations?
No, there are no seat limitations for the "Business" and "Enterprise" plans. The "Professional" plan is recommended for individual analysts, traders, influencers, and users and is therefore restricted to 1 seat.
19. What is a Carbon Arc MCP token?
Carbon Arc MCP tokens govern usage of Carbon Arc's MCP Server either natively through Lenses or non-natively via your preferred LLM.
MCP tokens are dedicated for use with Carbon Arc's MCP Server and cannot be used to purchase frameworks in Builder or programmatically. They are tracked separately from platform tokens.
20. What is the difference between a Carbon Arc MCP token and a Carbon Arc primary token? Are they 1:1?
There are several key differences between Carbon Arc MCP tokens and Carbon Arc primary tokens:
Usage Context: Carbon Arc MCP tokens are used to access Carbon Arc data through the Carbon Arc MCP Server via Lenses or your preferred LLM (e.g., ChatGPT or Claude). Carbon Arc primary tokens are used to purchase frameworks through the Builder or programmatically via the API.
Issuance and Expiration: Carbon Arc MCP tokens are allocated based on your subscription plan and reset daily at 12:00 AM Eastern Time. Carbon Arc primary tokens do not expire.
Consumption and Cost: While both token types query the same underlying data using the same structured framework objects and are priced using the same $/MB rate, they are not 1:1. The number of tokens consumed may differ depending on how an LLM interprets and constructs the framework object (e.g., entities, insights, time, and space filters). As a result, a query such as "Show me Walmart credit card spend in 2025" may consume a different number of tokens when executed via the Carbon Arc MCP Server compared to purchasing a similar framework through the Builder. In theory, if the identical structured framework object returned in a tool call result via the Carbon Arc MCP Server was used to buy data programmatically or in the Builder using the Carbon Arc primary token, the cost would be the same.
21. Are there usage limits?
Each subscription includes a daily MCP token allotment that automatically resets every 24 hours at 12:00am Eastern Time. Unused tokens will expire at the time of refresh. Once the daily limit is reached, additional usage can continue through purchased Primary MCP tokens. Primary MCP tokens do not expire and are consumed only after the included daily allotment has been exhausted.
22. What happens if I hit my daily token limit?
If your daily MCP token limit is reached, you may:
- Purchase Primary MCP tokens to continue using the Carbon Arc MCP Server
- Upgrade your plan (e.g., from "Professional" to "Business") via the Subscription page in the user portal
- Contact support@carbonarc.co to upgrade your plan as an Enterprise user
- Wait until your daily token limit resets at 12:00am Eastern Time to resume using the Carbon Arc MCP Server
23. What is the difference between my daily token limit and the Primary MCP token?
Daily token limits are automatically assigned based on the subscription plan purchased. These tokens are consumed before any Primary MCP tokens are consumed. Daily tokens expire at 12:00am Eastern Time and are reset daily. Primary MCP tokens never expire.
24. Who pays for the LLM tokens required to use the Carbon Arc MCP Server?
When using Lenses, Carbon Arc covers the LLM token costs associated with querying the Carbon Arc MCP Server through a self-hosted model. When accessing the Carbon Arc MCP Server through a non-native interface (e.g., ChatGPT or Claude), users are responsible for the LLM token costs associated with their own LLM account.
25. Do I need to have my own Claude or ChatGPT subscription to use the Carbon Arc MCP Server?
When using Lenses, clients do not need their own Claude or ChatGPT subscription to access the Carbon Arc MCP Server. When accessing the Carbon Arc MCP Server through a non-native interface (e.g., ChatGPT or Claude), clients must have their own LLM account.
26. Are Carbon Arc MCP tokens and LLM tokens the same?
No, Carbon Arc MCP tokens are used to measure and manage usage of the Carbon Arc MCP Server when querying Carbon Arc's data assets. LLM tokens are used by foundation models (e.g., Claude or ChatGPT) to process prompts and generate responses. MCP tokens govern access to Carbon Arc's data, while LLM tokens govern model computation.
27. Do all Carbon Arc MCP Server tool calls consume Carbon Arc MCP tokens?
No, not all MCP tool calls consume Carbon Arc MCP tokens. For example, discovery tools like "Search Entities" do not consume Carbon Arc MCP tokens. Analytical tools like the "Text to Insight" or "Entity Research" tool do.
28. How do I tell how many Carbon Arc MCP tokens I consumed with a query?
Lenses: the price for each MCP tool call is displayed natively in the UI.
Non-natively: this depends on the LLM provider selected. However, most LLMs including Claude and ChatGPT display a tool call "result". If you expand the result and scroll to the bottom, the price, available balance, and limit is returned.
29. Where can I see how many daily tokens I have remaining?
Lenses: daily token consumption and limits are displayed in the UI for ease of access. This is also available on the "Wallet" page in the User Portal.
Non-natively: this depends on the LLM provider selected. However, most LLMs including Claude and ChatGPT display a tool call "result". If you expand the result and scroll to the bottom, the price, available balance, and limit is returned.
30. How are Carbon Arc MCP tokens priced?
Carbon Arc MCP token usage is based on the volume and structure of data returned by the Carbon Arc MCP Server in response to a query. Queries that request broader time ranges, more entities, or more granular outputs will typically consume more MCP tokens than narrower, aggregated requests. To estimate usage in advance, limit the scope of your query (e.g., date range, geography, entity set) and request higher-level summaries where possible.
31. Why can't I calculate how many Carbon Arc MCP tokens will be consumed before I submit the prompt?
Because the exact size and composition of the returned data asset (e.g., number of entities, time series density, applicable filters) are not known until the query is executed, MCP token usage cannot be precisely calculated in advance.
This is not unique to the Carbon Arc MCP Server. Any MCP Server or LLM-driven data retrieval system must first resolve a natural language prompt into a structured query and execute it against underlying data assets before determining how much data will be returned. As a result, token consumption in MCP-based workflows is inherently dependent on the output of the query rather than the input prompt alone.
32. Can I re-run a previous query without consuming additional Carbon Arc MCP tokens?
No, since every prompt, orchestration, and result can be different you cannot re-run a previous query (even if identical) without consuming additional Carbon Arc MCP tokens.
33. What happens if a query fails or times out?
Carbon Arc MCP tokens should not be consumed if a query fails or times out. If you believe this has occurred, please reach out to support@carbonarc.co.
34. Should users expect any differences between outputs generated through Lenses and those generated by accessing the Carbon Arc MCP Server directly (e.g., through an LLM like Claude or ChatGPT)?
Outputs from Lenses and from a client directly accessing the Carbon Arc MCP Server may not be identical in all cases. While both rely on the same underlying MCP Server, Lenses applies Carbon Arc–defined orchestration, prompting frameworks, and workflow guardrails, whereas direct MCP usage within tools like Claude or ChatGPT depends on the client's own prompting, agent configuration, and potential additional reference material uploaded by the user. Any differences in output would therefore stem from differences in agent orchestration and user interaction design, not from differences in underlying data sources or platform governance.
35. Can clients upload documents or connect external files to the Carbon Arc MCP Server?
Lenses: No. Lenses interaction is limited to text-based prompts within defined character limits. It does not support document uploads or connections to client file repositories, shared drives, or other external systems.
Non-natively: Yes, depending on which LLM provider is used, documents or other reference material may be uploaded or attached in the user's own instance.
36. Are Carbon Arc clients able to give Lenses access to any of the user's other reference materials (e.g., a powerpoint with a stock thesis, access to a shared drive on their side, etc.) that they are able to give Claude access to when using MCP outside the platform?
No. Carbon Arc clients cannot grant Lenses access to external reference materials such as internal documents, shared drives, or other systems. Interaction with Lenses is limited to text-based prompts within defined character limits, and the feature does not support document uploads or connections to client file repositories or infrastructure.
37. How are responses to prompts derived by the Carbon Arc MCP Server?
Responses are derived from three sources:
- The user's submitted prompt
- Structured data accessed through the Carbon Arc MCP Server
- Processing by a foundation model that summarizes context rather than introducing additional data sources
38. Will Lenses outputs differ from those generated via non-native access to the Carbon Arc MCP Server?
Outputs may not be identical. While both use the same underlying Carbon Arc MCP Server and data, Lenses applies Carbon Arc–defined orchestration, prompting frameworks, and workflow guardrails. Non-native MCP usage depends on the client's own prompting, agent configuration, and any additional reference materials they provide. Differences stem from interaction design—not from differences in underlying data.
39. Will the LLM cite the underlying Carbon Arc data used to create output?
The LLM can be prompted to cite the Insight ID and Topic (for example, Insight ID 775 – "App Downloads", CA0054 – Core Panel), but it may or may not choose to cite the underlying data automatically. However, it will be clear when data is sourced from Carbon Arc because the MCP tool calls are returned.
40. What do I do if I want to get access to the raw data underlying the output from the Carbon Arc MCP Server?
The Carbon Arc MCP Server returns the tool call result which contains the structured framework object leveraged in the response. Users can use this framework object to pay for the data on the Carbon Arc platform either by accessing the Builder through the web application or programmatically via the Carbon Arc SDK.
41. What does the Carbon Arc MCP Server provide in response to a prompt?
The Carbon Arc MCP Server can return data or summarized insights depending on the orchestration of the agent.
42. Will two users submitting the same query receive the same output?
No, the Carbon Arc MCP Server is reliant on model orchestration which may differ.
43. Can I customize the format of the output?
Lenses: this is available with limited functionality based on formatting prompts.
Non-natively: yes, you can prompt the LLM to customize the formatting or output as you would prefer. For example, ask Claude to "build an interactive dashboard artifact to visualize the data and response."
44. Where can I see a history of transactions?
Carbon Arc MCP Server transactions are recorded in a "ledger" format in the "Wallet" page of the Carbon Arc web application. Users can also see previous chat history in Lenses and non-natively.
Data Handling, Privacy, & Compliance
45. Can Carbon Arc see the prompts clients ask the Carbon Arc MCP Server?
Lenses: Yes, Carbon Arc is able to see the prompt clients ask.
Non-natively: No. Carbon Arc can only see what the LLM chooses to pass to each Carbon Arc tool under the Model Context Protocol. The LLM may adjust, simplify, or change the original prompt before passing it along. Only the synthesized portion of the prompt necessary to generate the response is shared. Clients can view exactly what was visible to Carbon Arc by expanding the tool box displayed by the LLM.
46. Are user prompts ever shared with third parties or surfaced in a way that other customers might see?
No. Prompts shared with Carbon Arc are not surfaced to third parties or made visible to other customers.
47. What information does Carbon Arc retain from Carbon Arc client queries?
Carbon Arc retains client-submitted queries (synthesized for non-native access) and limited associated operational metadata, such as session identifiers, timestamps, user email, feature usage, and system processing metrics (e.g., token spend).
48. How does Carbon Arc store the information it retains from inputs and/or outputs? Who has access to it?
Carbon Arc stores retained query and response data, along with associated operational metadata (e.g., user email, session identifiers, created_at timestamps, updated_at timestamps). This data is stored within Carbon Arc's internal databases under our standard platform security controls, with access restricted to authorized personnel with credentialed access.
Retained data is used solely for platform operations and is not used to train any AI foundation models.
49. What does Carbon Arc do with the information it retains from inputs and/or outputs (if any)?
Carbon Arc retains query and response data solely for platform operations such as system reliability, debugging, monitoring, and support. This information is not used to train foundation models, develop generalized model improvements, or run evaluation programs beyond operational troubleshooting and performance monitoring.
50. What agreement or terms of use/service governs Carbon Arc's Claude account?
Carbon Arc's use of Claude models via AWS Bedrock is governed by the AWS Customer Agreement and the AWS Bedrock Terms (including the Bedrock section of the AWS Service Terms), since access is provided through AWS rather than by Anthropic directly. Any model-specific obligations (e.g., acceptable use, data handling) are those incorporated into or referenced by AWS's Bedrock legal terms, not a standalone Claude plan from Anthropic. The precise governing terms can be found in the AWS Console under Bedrock legal links or in your executed AWS agreement (e.g., enterprise/EDP contract and the AWS Service Terms "Bedrock" section).
51. Do you take the view that Lenses should be considered Carbon Arc's generative AI model (i.e., "powered by Claude", but Carbon Arc's product), or that Lenses would still be considered to be a Carbon Arc client's instance of their own Claude?
Lenses would generally be reviewed as an additional AI-enabled feature within the existing approved Carbon Arc platform. It operates using the same Carbon Arc MCP Server, governance controls, and data flow boundaries already assessed, it does not represent a new model environment or external AI tool. Lenses does provide Carbon Arc visibility into a users prompt verbatim, but does not allow users to upload any documents.
52. Will the information Carbon Arc clients put into a Lenses query be used to train any generative AI model – Lenses and/or the underlying Claude model?
No. When Lenses uses AWS Bedrock to access Claude or other foundation models, client inputs and model outputs are not used to train those underlying models. Amazon Bedrock states that it does not use customer prompts or completions to train AWS models or share them with third parties, and that customer inputs and outputs are not stored for model training unless explicitly opted into by the customer.
53. What if I am an Enterprise client that is "compliance managed" and has restrictions enabled for my organization?
The Carbon Arc MCP Server respects compliance restrictions and will not surface data or insights from that data if the data asset is restricted for the Enterprise organization.
54. Are Carbon Arc clients subject to Anthropic's Acceptable Use Policy ("AUP")?
Yes. Carbon Arc's Terms of Service require all clients and their authorized users to comply with Anthropic's AUP, which applies to all use of AI-powered features, including Lenses and the Carbon Arc MCP Server. The AUP is available at https://www.anthropic.com/legal/aup.
The AUP is organized into three tiers: Universal Policies (applicable to all uses), High-Risk Use Case requirements (applicable to specific domains including Finance and Legal), and Additional Policies.
Clients whose use cases fall within the Finance or Legal categories should review the High-Risk Use Case requirements carefully, as additional obligations apply. Carbon Arc reserves the right to suspend or terminate access for AUP violations.
55. What are "High-Risk Use Cases" and do they apply to me?
Anthropic's AUP designates certain use cases as "High-Risk" and imposes additional requirements on users operating in those categories. Finance (including investment research, portfolio analysis, and trading-related applications) and Legal are both listed High-Risk Use Case categories.
If your use of Lenses or the Carbon Arc MCP Server constitutes a High-Risk Use Case, Carbon Arc's Terms of Service require you to:
- Ensure that a qualified human professional reviews AI-generated outputs before those outputs are disseminated or acted upon in a material way
- Disclose to your end users at the start of each session that outputs are generated with the assistance of AI
- Provide Carbon Arc with a written representation of whether your use constitutes a High-Risk Use Case
- These requirements reflect Carbon Arc's obligations as an operator under the Anthropic AUP and are conditions of access to AI-powered features.
56. Are there geographic restrictions on access to Lenses and Carbon Arc's MCP Server?
Yes, with an important clarification: Anthropic's Supported Regions Policy restricts access to its AI infrastructure based on entity ownership and control — not based on where individual users are physically located.
Specifically:
- Carbon Arc's clients are responsible for ensuring that neither they nor any of their affiliates is an entity whose majority ownership or effective control is attributable, directly or indirectly, to a jurisdiction not included on Anthropic's Supported Regions list (available at https://www.anthropic.com/supported-countries).
- Clients must also ensure they do not extend access to Lenses or the MCP Server to any sub-licensed entity, fund, or affiliate that would itself be ineligible under the Anthropic Supported Regions Policy.
- Clients who are in breach of this representation risk suspension of their access to AI-powered features, as a breach may jeopardize Carbon Arc's ability to maintain its relationship with Anthropic.
This restriction does not affect access to Carbon Arc platform features that do not involve AI processing. The Supported Regions list is maintained by Anthropic and is subject to change. For questions about whether a specific entity structure implicates this restriction, please contact compliance@carbonarc.co.
57. Can I submit personal information about individuals in my queries?
No. Carbon Arc's Terms of Service prohibit clients from submitting personal information identifying natural persons in queries to AI-powered features, including Lenses and the Carbon Arc MCP Server. This includes names, identification numbers, account numbers, investor details, portfolio holder information, and any other information that identifies or could be used to identify an individual.
This restriction applies to both Lenses (native access) and non-native MCP Server access. Queries should be limited to entity-level, market, or industry data. If you have questions about whether specific data categories are permissible, please contact support@carbonarc.co.