Credal

What is Credal?

Credal is an enterprise agent control plane that turns team knowledge and reusable skills into governed AI capabilities. Organizations can connect business data, compose specialized agents, publish skills to a registry, and expose them through team-scoped MCP servers, chat products, or application APIs.

The platform inherits permissions from connected systems and synchronizes identity, groups, and roles through an identity provider. A user querying an agent can only retrieve information they already have permission to access. Teams can version skills, control which actions an agent may call, require human approval for sensitive operations, and inspect one consolidated audit trail across agent surfaces.

Credal is model-independent and can apply the same connector, prompt, permission, and approval behavior across chat assistants, coding tools, Slack, and custom applications. Deployment options include Credal's cloud or an organization's own VPC.

How to Use Credal

1. Connect the organization's identity provider and required data sources.
2. Import or build a skill using approved business context and actions.
3. Define which teams, groups, or individuals may discover and invoke it.
4. Add action scopes and human approval gates for sensitive operations.
5. Publish the skill to the registry or a scoped MCP server.
6. Deploy it to supported chat surfaces or applications and monitor the audit trail, usage, and spend.

Core Features

• Governed skill registry: Versions and shares reusable agent capabilities across teams.
• Permission-aware grounding: Inherits source permissions and keeps access synchronized with SSO.
• Scoped MCP servers: Gives each team an MCP surface containing only its approved skills and data.
• Cross-surface consistency: Applies the same behavior across models, chat tools, IDEs, and apps.
• Action controls: Restricts tools and adds human approval before sensitive execution.
• Central auditing: Consolidates queries, approvals, actions, spend, and agent activity.
• Flexible deployment: Supports a secure hosted environment or deployment inside a customer VPC.

Use Cases

• Internal knowledge agents: Answer questions using permission-filtered company data.
• Reusable department skills: Publish a proven workflow once instead of rebuilding it for each team.
• Enterprise MCP governance: Register and control the MCP tools available to users and agents.
• Agent action security: Require approval before writing messages, updating records, or creating tasks.
• Multi-model operations: Change model or interface without recreating permissions and connectors.

Pricing

Credal provides a sales-led demo and a pricing page, but the public product page does not expose a simple, verifiable self-service price. Organizations should request a current quote based on deployment and governance requirements.

Frequently Asked Questions

Does Credal replace the underlying AI model?

No. It is a governed layer under multiple models and interfaces.

How does it prevent data over-sharing?

Connected-source permissions are inherited and synchronized, and each query runs with the requesting user's allowed access.

Can sensitive actions require review?

Yes. Teams can add human approval gates and keep a record of who approved an action.