KeyRunner builds a governed execution layer for enterprise AI agents. Instead of handing agents API keys, it turns enterprise APIs into named actions that agents can call, injects credentials at runtime from the secret stores a company already runs, and checks every call against policy before it reaches the target system.
The platform is aimed at platform engineers and security teams who want agents such as Claude, ChatGPT and Cursor to work with business systems without exposing secrets to prompts or agent code.
At A Glance
| Vendor | KeyRunner (LaunchIAM Inc.) |
|---|---|
| Category | Secrets handling and governed execution for AI agent tool calls |
| Identities covered | AI agents, developers and API credentials held in secret stores |
| Deployment | Desktop app, VS Code extension and npm CLI; AI tooling runs inside the customer environment |
| Agentic AI | Governed API tools for agents, served over MCP |
| Founded | 2023, Austin, Texas |
| Market classification | Secrets Management; also Agentic AI Identity |
What KeyRunner Does
The principle behind the product is short: “Agents should get actions, not API keys.” KeyRunner places one governed layer between agents and enterprise systems. Agents see only an approved set of tools, credentials stay inside the customer’s infrastructure, and secrets are resolved at the moment a call executes. The same approach starts with developers testing requests locally and carries through to agents running in production. Security teams get one place to decide which actions each agent may take, and one record of what it actually did.
Products
API Client
The API Client lets developers build and test API requests on their own machine, with no account needed. Requests run locally and nothing is logged by default, which suits teams that do not want credentials passing through a third-party service. Secrets are pulled from Vault, 1Password or AWS Secrets Manager at request time, so they never land in collections or request history. Mock servers let developers simulate an API before the real endpoint exists. It is available for Windows and Mac, as a VS Code extension and as an npm CLI.
Agent Security
Agent Security is for organisations putting agents into production that need to control what those agents are allowed to do. Every tool call is governed before it runs, using role-based tool access and policy as code, and credentials are injected from the vault rather than given to the agent.
- Per-tool circuit breakers stop execution automatically when a tool shows anomalous patterns
- Each run produces an execution ID and an audit record
- Secret resolution is logged without recording the secret values
AI Tools Add-On
The AI Tools add-on converts existing APIs into tools that agents can use, served through a KeyRunner MCP server. Raw API keys stay away from the agent, sensitive data is redacted from responses, and each tool call leaves an audit record. The same tools can be offered to OpenAI function calling, with the same policy checks. In the AI workspace, teams turn tested requests into governed tools, compose policies, connect production agents and inspect individual executions.
How It Works
Every agent tool call follows a fixed pipeline inside the customer’s infrastructure boundary.
| Component | What it does |
|---|---|
| Secret resolution | Environment and secret store references are resolved when the call executes, so the agent never holds the key. |
| Policy check | The call is evaluated against policy first; blocked calls never reach the target API. |
| Response handling | Responses are filtered for PII, PHI, PCI data and secrets, and rate limits apply. |
| Observability | The desktop app provides searchable traces of model calls, policy decisions, tool calls, API responses, latency, token use, cost and failures. |
Identities And Systems Covered
| AI agents | Claude, ChatGPT, Cursor and internal agents calling governed tools. |
|---|---|
| Credentials and secrets | API credentials held in HashiCorp Vault, AWS Secrets Manager, Azure Key Vault, GCP Secret Manager and 1Password Secrets Automation. |
| Developers and teams | Human users managed with SSO and RBAC on the Explorer plan. |
| Enterprise systems | Salesforce, ServiceNow, Jira, Workday, Snowflake and any REST API. |
Integrations
| Area | Integrations |
|---|---|
| AI frameworks and coding tools | OpenAI function calling, LangChain, CrewAI, Claude Code |
| CI/CD | GitHub Actions, GitLab CI |
| Developer tools | OpenAPI and Swagger, Postman Collections |
| API management | Apigee, AWS API Gateway, Kafka |
| Audit and monitoring | Splunk, Datadog, Slack, generic SIEM or webhook |
Company Background
| Company | KeyRunner is built by LaunchIAM Inc. |
|---|---|
| Founded | Incorporated in 2023. |
| Headquarters | Austin, Texas, United States. |
Note: This overview is compiled from KeyRunner’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.