diff --git a/.cspell.json b/.cspell.json index 420dd6e..eca61a0 100644 --- a/.cspell.json +++ b/.cspell.json @@ -78,7 +78,7 @@ "webkitallowfullscreen", "mozallowfullscreen", "Cowork", - "automagically" + "blocklists" ], "flagWords": ["localhost", "127.0.0.1"], "ignorePaths": ["node_modules", ".git", "dist", "build", "*.log"], diff --git a/fern/docs/assets/actions/custom-name-desc.png b/fern/docs/assets/actions/custom-name-desc.png new file mode 100644 index 0000000..9ba07b7 Binary files /dev/null and b/fern/docs/assets/actions/custom-name-desc.png differ diff --git a/fern/docs/assets/actions/monitor-actions.png b/fern/docs/assets/actions/monitor-actions.png new file mode 100644 index 0000000..688ed39 Binary files /dev/null and b/fern/docs/assets/actions/monitor-actions.png differ diff --git a/fern/docs/pages/administration/connect-mcp-clients.mdx b/fern/docs/pages/administration/connect-mcp-clients.mdx index 5d90bb7..fd64e07 100644 --- a/fern/docs/pages/administration/connect-mcp-clients.mdx +++ b/fern/docs/pages/administration/connect-mcp-clients.mdx @@ -4,7 +4,7 @@ slug: user-guide/platform/administration/connect-mcp-clients subtitle: Curate which MCP Servers and Agents appear in which of your AI chat surfaces --- -Apps such as ChatGPT, Claude, Claude Cowork or Cursor, allow administrators to **push out MCP servers** to their users: **add Credal's MCP Discovery URL** to automatically push your organization's MCP servers and agents to these apps. Then, when users enable "Claude" or "Cursor" on their preferred MCP servers those apps can automagically connect. +Apps such as ChatGPT, Claude, Claude Cowork or Cursor, allow administrators to **push out MCP servers** to their users: **add Credal's MCP Discovery URL** to automatically push your organization's MCP servers and agents to these apps. Then, when users enable "Claude" or "Cursor" on their preferred MCP servers those apps can automatically connect. **Apps connect to Credal via OAuth.** This works out of the box with no per-app configuration required. diff --git a/fern/docs/pages/introduction.mdx b/fern/docs/pages/introduction.mdx index b6b8158..8f031f4 100644 --- a/fern/docs/pages/introduction.mdx +++ b/fern/docs/pages/introduction.mdx @@ -20,47 +20,72 @@ title: Introduction -## What is Credal? +# What is Credal? -Credal is a platform that empowers enterprises to build secure, intelligent AI agents tailored to any workflow, with native support for enterprise data and access controls. +Credal gives enterprises a library of hosted agents and tools with built-in controls, accessible wherever your employees want to chat with AI. It's the centralized trust layer for governing and scaling AI agents, MCP servers, and skills—connected to any system, operating within the security and permission boundaries your organization requires. -## What is an Agent? +## What Can You Build? -An agent is more than just a chatbot. It's a system capable of reasoning and taking actions to achieve goals. It combines two core capabilities: +Credal supports the full stack of agentic infrastructure: -1. **Reasoning** – Analyzing information, context, or instructions to determine the next steps. -2. **Action-taking** – Performing tasks such as calling APIs, generating outputs, or interacting with external systems based on its reasoning. +- **Agents** – Systems that reason and take action to complete goals across your enterprise data and systems. +- **MCP Servers** – Expose governed, curated collections of tools and data that any agent or AI application can call. +- **Skills** – Reusable capabilities that agents can invoke—queryable knowledge, API calls, workflows—built once and available everywhere. -Unlike simple chatbots that just respond, agents **reason and act**. For example: - -- A chatbot might answer: “It’s 75°F and sunny in New York.” -- An agent could fulfill: "Book me a flight to New York for tomorrow," by searching for flights, choosing the best option, and completing the booking. +Everything you build in Credal is governed from day one: permissioned, auditable, and deployable at enterprise scale. ## Why Credal? -### 1. **Enterprise Data Integrations** +### A Governed Library of Agents and Tools + +Credal provides a central library of tools and company data collections that your AI platform team can customize, document, and govern. Tools can be enriched with curated context, scoped to the right audiences, and extended dynamically—Credal can write code to perform complex multi-step workflows and repackage them as first-class tools in your library. Builders and users can browse, activate, and stack vetted tools without starting from scratch. + +### Agents That Work Where Your Teams Work + +Credal publishes agents to every surface your teams already use—Slack, the web, APIs, chat interfaces, and more. Users get access to agentic tooling where they work, not in a separate tool. And as those surfaces multiply, you need a single place where tooling is created, maintained, and governed—otherwise you're managing sprawl. + +### An Agent Builder with Built-In Evaluation + +Build agents through a UI, in natural language, or via API. Credal's agent builder includes native evaluation capability so you can test, iterate, and validate before deploying to your organization—closing the gap between prototype and production. + +### Model-Agnostic by Design + +Credal works across model providers. Most enterprises already use multiple LLMs because different models excel at different tasks. Credal lets you build agentic systems that use the right model for the right job, without rebuilding for each provider. + +### Market-Leading Enterprise Context -Connect to all major [enterprise data sources](/user-guide/platform/integrations/data-sources-action-providers)—Slack, Confluence, Google Drive, Salesforce, databases, and more—for unified search, knowledge retrieval, and advanced analytics. With enterprise-grade security, every API connection and query follows your organization's access controls, ensuring your agents can safely surface insights and documents across your company's knowledge base. Whether answering questions, driving business processes, or enabling complex workflows, your data remains fully protected. +Index and curate unstructured data—Slack, Confluence, Google Drive, Salesforce, databases, and more—and make it available to any agent or AI application from a single place. Structured and semi-structured data sources can be centrally configured and fetched securely. Your agents get better context; your admins get full visibility into what's being accessed. -### 2. **Agentic Capabilities** +### Administrative Visibility and Control -Build intelligent agents that collaborate to execute multi-step workflows across your enterprise data. Agents automatically share capabilities—when one agent learns to query Salesforce, every agent in your organization can leverage that skill. Design sophisticated, secure workflows where [specialized agents coordinate](/user-guide/platform/agent-builder/configuration/agent-collaboration) on complex tasks, accelerating decision-making while maintaining strict data governance. +Credal isn't just secure—it's where you can safely deploy agents that take real actions: sending emails, processing payments, modifying records. -### 3. **Publish Anywhere** +**For admins:** +- A single control panel for all agents and tools across the organization +- See every agent using a given system (e.g., "What are all the agents using Google Drive for?") +- Identify documents with incorrect underlying permissions before AI surfaces them further +- Set global rules: require approval for external emails, cap payment amounts, block destructive operations +- Enforce permission boundaries across every agent in the enterprise (authentication, authorization, auditability) +- Measure and drive adoption across the user base from a single dashboard -Publish your agents wherever your teams work. Use them on the web, in Slack, through APIs, with other agents, or in any chat interface (with more integrations coming soon). Every published agent inherits Credal's enterprise-grade security, ensuring your knowledge remains private while your teams stay productive. +**For builders:** +- Browse and activate vetted connectors, skills, and agents from a central hub +- Evaluate solutions and scale them to teams and the organization quickly +- Stack governed, pre-vetted tools to accelerate time-to-production +- Benefit from better search and response accuracy from curated, well-scoped tooling ## Security -Credal is built for enterprises—designed to meet the security, compliance, and governance standards required by large organizations. +Credal is built for enterprises—designed to meet the security, compliance, and governance standards large organizations require. - SOC 2 Type II certified, HIPAA-ready, and compliant with the EU/UK/Swiss Data Privacy Framework -- Credal mirrors your existing permissions (Slack, Google, Salesforce, etc.), so agents only access what users are already authorized to see -- Supports human-in-the-loop approval flows and detailed [audit logs](/user-guide/platform/administration/audit-log-data-exports) -- Flexible deployment: Cloud or on-prem, with [SSO and SCIM](/user-guide/platform/administration/single-sign-on-sso) integration for identity and access management +- Mirrors your existing permissions (Slack, Google, Salesforce, etc.)—agents only access what users are already authorized to see +- Supports human-in-the-loop approval flows and detailed audit logs +- Flexible deployment: cloud or on-prem, with SSO and SCIM integration for identity and access management +- Global policy controls for how agents interact with external systems, enforced across every agent in your organization -With Credal, enterprises can confidently deploy AI agents that accelerate knowledge workflows without ever compromising security or compliance. +With Credal, enterprises can confidently move from AI experimentation to production—without compromising security, compliance, or control. ## Support -For questions or support? Reach out to our team at **[support@credal.ai](mailto:support@credal.ai)**. +Questions or need help? Reach out at **support@credal.ai**. diff --git a/fern/docs/pages/platform/actions/action-human-approval.mdx b/fern/docs/pages/platform/actions/action-human-approval.mdx index 673d040..1dcdb08 100644 --- a/fern/docs/pages/platform/actions/action-human-approval.mdx +++ b/fern/docs/pages/platform/actions/action-human-approval.mdx @@ -23,3 +23,12 @@ An operation is a multi-user operation if the approval to carry out a given acti - All Actions in an Agent that undergo permission escalation must have the Action Release Gate turned on. If a builder wants to turn it off for whatever reason, there must be a justification provided with that toggle off. The ARG will require an egress approval step from the Agent owner(s). This way, humans can use their expertise and judgment to ensure DLP. All three of these provisions together make your multi-user workflows airtight and prevent any unauthorized access, visibility, or indirect leakage. + +## Coming Soon + +We're actively developing additional human approval capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Organization-level HITL enforcement**: Configure human approval requirements at the organization level to ensure consistent governance across all actions +- **HITL for MCP server actions**: Extend human-in-the-loop approval flows to actions executed through imported MCP servers +- **Multi-user approval workflows with permissions escalation**: Advanced workflows where multiple approvers can provision temporary or scoped access to resources, enabling complex cross-functional operations with granular permission grants +- **Version control and review workflows**: GitHub-style review process for changes to agents, actions, and skills with branching, change approval from multiple reviewers, and version control before deploying updates diff --git a/fern/docs/pages/platform/actions/action-parameters.mdx b/fern/docs/pages/platform/actions/action-parameters.mdx index 15f9892..7bd8355 100644 --- a/fern/docs/pages/platform/actions/action-parameters.mdx +++ b/fern/docs/pages/platform/actions/action-parameters.mdx @@ -50,3 +50,11 @@ parameters: + +## Coming Soon + +We're actively developing advanced parameter capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Conditional HITL rules**: Define parameter-based rules that trigger human approval only when specific conditions are met (e.g., require approval for expenses over $10,000) +- **Conditional execution rules**: Create parameter-based logic that determines whether an action should execute based on specific criteria +- **Runtime context parameters**: Automatically inject contextual information at execution time, such as current timestamp, triggering event details, or dynamic environment variables diff --git a/fern/docs/pages/platform/actions/action-security-model.mdx b/fern/docs/pages/platform/actions/action-security-model.mdx index 5f7539f..8501694 100644 --- a/fern/docs/pages/platform/actions/action-security-model.mdx +++ b/fern/docs/pages/platform/actions/action-security-model.mdx @@ -58,3 +58,10 @@ Separately from respecting the permissions of a user in an action provider's und In other cases, it may be desirable for users to kick off an action invocation despite not having the permissions in the underlying system to execute that action. Using the human approval flow, it's possible to request for someone with those permissions to approve the action and execute it on the original user's behalf. Learn more about this flow: [Human Approval Options](/user-guide/platform/governed-actions/human-approval-options) + +## Coming Soon + +We're actively developing additional security and governance capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Role-based action access**: Grant access to actions based on user roles, allowing fine-grained control over who can invoke specific actions +- **Action blocklists**: Define organization-wide policies to prevent specific actions from being used in certain contexts or by certain user groups diff --git a/fern/docs/pages/platform/actions/custom-names-descriptions.mdx b/fern/docs/pages/platform/actions/custom-names-descriptions.mdx new file mode 100644 index 0000000..0d02ca4 --- /dev/null +++ b/fern/docs/pages/platform/actions/custom-names-descriptions.mdx @@ -0,0 +1,36 @@ +# Customizing Actions + +Credal allows you to assign custom names and descriptions to actions, making them more intuitive and aligned with your organization's business logic and terminology. + +## Why Customize Action Names and Descriptions? + +Third-party integrations and MCP servers often expose actions with technical names that may not reflect how your team thinks about these operations. Custom names and descriptions help: + +- **Align with Business Terminology**: Use names that match your organization's vocabulary and processes +- **Improve Discoverability**: Make actions easier to find and understand for team members +- **Enhance Agent Performance**: Clear, business-focused descriptions help AI agents select the right actions more accurately +- **Simplify Onboarding**: New team members can understand what actions do without learning technical jargon + +## How to Customize + +When configuring an action, you can override the default name and description: + +1. Navigate to the action you want to customize in the Actions section +2. Click on the action to open its settings +3. Edit the **Name** field to provide a custom, business-friendly name +4. Update the **Description** field to explain what the action does in terms relevant to your organization + +![custom-name-desc.png](/docs/assets/actions/custom-name-desc.png) + +## Best Practices + +- **Be Descriptive**: Use clear, concise names that immediately convey the action's purpose +- **Include Context**: Add business context in descriptions (e.g., "Creates a Jira ticket in our Support queue" instead of "Create Jira issue") +- **Stay Consistent**: Use consistent naming conventions across similar actions +- **Think About Discovery**: Consider how users will search for this action when naming it + + +Custom names and descriptions only affect how actions appear within Credal. The underlying integration or MCP server functionality remains unchanged. + + +You can distribute and scale usage of your customized actions in any surface through our [custom MCP servers](https://docs.credal.ai/user-guide/platform/mcp-servers/overview). Your HR, Finance, Product, etc. team can then manage their own blessed set of tools for the entire organization to use. diff --git a/fern/docs/pages/platform/actions/monitoring.mdx b/fern/docs/pages/platform/actions/monitoring.mdx index 3e2452c..5d18da6 100644 --- a/fern/docs/pages/platform/actions/monitoring.mdx +++ b/fern/docs/pages/platform/actions/monitoring.mdx @@ -1 +1,23 @@ # Monitoring Usage + +Credal provides comprehensive monitoring capabilities to track how actions are being used across your organization. The monitoring dashboard gives you visibility into action invocations, execution patterns, and usage trends. + +![monitor-actions.png](/docs/assets/actions/monitor-actions.png) + +## What You Can Monitor + +- **Action Invocations**: See how often each action is being called by agents +- **Execution Success Rates**: Track which actions are completing successfully vs. failing +- **User Activity**: Understand which users and teams are leveraging different actions +- **Action Justifications**: Review business justifications provided by users when invoking sensitive actions, creating a complete audit trail for compliance and governance +- **Temporal Patterns**: Identify usage trends over time to optimize action configurations + +## Accessing Monitoring Data + +To view action monitoring data: + +1. Navigate to the Actions section in the Credal platform +2. Select the action you want to monitor +3. Click on the **Monitoring** or **Usage** tab + +The monitoring dashboard provides real-time insights to help you optimize action performance and governance. diff --git a/fern/docs/pages/platform/actions/overview.mdx b/fern/docs/pages/platform/actions/overview.mdx index c7f1a14..c40561f 100644 --- a/fern/docs/pages/platform/actions/overview.mdx +++ b/fern/docs/pages/platform/actions/overview.mdx @@ -25,3 +25,9 @@ To get started: **Note:** You can create custom actions to interface with your private internal systems using the "Fetch from an API or trigger a webhook" template and pasting in an OpenAPI spec! Before we get to creating your first action, let's learn about [action parameters](/user-guide/platform/governed-actions/action-parameters)! + +## Coming Soon + +We're actively developing additional action management capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Rate limiting for actions**: Set usage limits for actions to prevent abuse and control costs, including per-user, per-agent, and organization-wide rate limits diff --git a/fern/docs/pages/platform/agents/agent-evaluate.mdx b/fern/docs/pages/platform/agents/agent-evaluate.mdx index d302f11..8fd3845 100644 --- a/fern/docs/pages/platform/agents/agent-evaluate.mdx +++ b/fern/docs/pages/platform/agents/agent-evaluate.mdx @@ -90,3 +90,9 @@ Watch a video walking you through how to use the Evaluate tab below: referrerpolicy="strict-origin-when-cross-origin" allowfullscreen > + +## Coming Soon + +We're actively developing additional evaluation capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Evaluation gate before publishing**: Require agents to pass evaluation thresholds before they can be published, ensuring quality standards are met before deployment to end users diff --git a/fern/docs/pages/platform/agents/agent-testing.mdx b/fern/docs/pages/platform/agents/agent-testing.mdx index 6084cb5..1b650c8 100644 --- a/fern/docs/pages/platform/agents/agent-testing.mdx +++ b/fern/docs/pages/platform/agents/agent-testing.mdx @@ -43,6 +43,17 @@ For a comprehensive guide on writing and optimizing agent instructions, see [Wri To evaluate prompt changes at scale and systematically measure their impact on agent performance, use [Evaluations](/user-guide/platform/agent-builder/evaluating-your-agent). This allows you to test different prompt variations against a set of test cases and compare results objectively. +## Testing Actions with Dry Run Mode + +When testing agents that use actions, you can use dry run/preview mode to see what an action would do without actually executing it. This is particularly useful for: + +- Testing actions that modify data (create, update, delete operations) +- Validating that the agent is selecting the right actions and parameters +- Reducing risk when testing new workflows in production environments +- Training and onboarding scenarios where you want to demonstrate functionality without side effects + +Dry run mode allows you to safely test your agent's decision-making and action invocation logic before allowing real execution. + --- ## Troubleshooting diff --git a/fern/docs/pages/platform/agents/deployment/overview.mdx b/fern/docs/pages/platform/agents/deployment/overview.mdx index 3e10bba..6ade9e9 100644 --- a/fern/docs/pages/platform/agents/deployment/overview.mdx +++ b/fern/docs/pages/platform/agents/deployment/overview.mdx @@ -57,3 +57,18 @@ After publishing, your agent will appear under **My Published** in the **Agents* - Published agents can be unpublished or updated anytime from the **Publish** tab. ![agent-draft-deploy.png](/docs/assets/agents/agent-draft-deploy.png) + +--- + +## Coming Soon + +We're actively developing advanced agent deployment and security capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Version control and review workflows**: GitHub-style review process for changes to agents with branching, change approval from multiple reviewers, and version control before deploying updates +- **Sensitive agents**: Mark agents as sensitive to apply additional security controls and restrict access to authorized users only +- **Automatic hallucination detection**: Real-time monitoring and flagging of potential hallucinations in agent responses +- **Automatic risk detection**: Intelligent identification of high-risk operations or outputs that may require additional review +- **Agent-to-agent action delegation controls**: Fine-grained permissions for how agents can delegate tasks and actions to other agents +- **Client-level access controls**: Control which applications, surfaces, or clients can access specific agents +- **Prompt injection prevention**: Advanced safeguards to detect and block adversarial prompt injection attempts +- **Templates per agent/use case**: Pre-configured agent templates optimized for specific business functions (HR, Finance, Product, etc.) with recommended settings and guardrails diff --git a/fern/docs/pages/platform/mcp-servers/import-server/setting-up-controls.mdx b/fern/docs/pages/platform/mcp-servers/import-server/setting-up-controls.mdx index 2bb62a7..ed76d94 100644 --- a/fern/docs/pages/platform/mcp-servers/import-server/setting-up-controls.mdx +++ b/fern/docs/pages/platform/mcp-servers/import-server/setting-up-controls.mdx @@ -1,3 +1,7 @@ # Setting up Controls For every MCP server, admins can enable or disable individual tools and require [human approval](/user-guide/platform/governed-actions/human-approval-options) before specific actions execute. If a third-party server exposes actions that are too sensitive or not relevant for your organization, simply turn them off. For actions you want to allow but monitor closely, approval enforcement ensures nothing runs without a human sign-off first. + + +**Tool Syncing:** Credal automatically refreshes the list of available tools every time `tools/list` is called through the MCP gateway. You can also manually refresh tools using the refresh button in the UI. Background refreshes with version control to synchronously approve changes are coming soon, as well as the ability to enforce human approval on any of these actions. + diff --git a/fern/docs/pages/platform/mcp-servers/overview.mdx b/fern/docs/pages/platform/mcp-servers/overview.mdx index e29d59e..4fdccf1 100644 --- a/fern/docs/pages/platform/mcp-servers/overview.mdx +++ b/fern/docs/pages/platform/mcp-servers/overview.mdx @@ -15,3 +15,11 @@ All MCP server activity is subject to Credal's access control and approval enfor ## Building with MCP Credal provides a broad library of [pre-built tools](/user-guide/platform/governed-actions/overview) that can be incorporated into your MCP servers, making it straightforward to connect data sources, trigger actions, and extend capability without building from scratch. + +## Coming Soon + +We're actively developing advanced MCP server management capabilities. If you'd like early access to any of these features, please contact us at [support@credal.ai](mailto:support@credal.ai): + +- **Organization-wide skills registry**: Centralized catalog of all approved tools, actions, and capabilities across your organization for easy discovery and reuse +- **MCP Server registry**: Comprehensive registry of all MCP servers deployed across your organization with versioning, dependencies, and usage tracking +- **Server and agent registry guardrails**: Automated governance controls and policies that ensure MCP servers and agents meet security, compliance, and evaluation standards before deployment diff --git a/fern/docs/user-guide.yml b/fern/docs/user-guide.yml index 487a89a..fd4953e 100644 --- a/fern/docs/user-guide.yml +++ b/fern/docs/user-guide.yml @@ -106,6 +106,8 @@ navigation: contents: - page: Overview path: ./pages/platform/actions/overview.mdx + - page: Custom Names and Descriptions + path: ./pages/platform/actions/custom-names-descriptions.mdx - page: Open Source Actions path: ./pages/platform/actions/open-source-actions.mdx - page: Credentials and Security