概览
What is Costwright Mcp?
Costwright Mcp is a static worst-case token‑budget analyzer for LLM‑agent workflows. It examines a Python repository that uses LangGraph, CrewAI, or the OpenAI Agents SDK and reports – by pure AST analysis, without running the code – the worst‑case budget ceiling of every workflow graph. It can optionally issue an Ed25519‑signed budget certificate that is logged to a public transparency log.
How to use Costwright Mcp?
Configure the MCP server using npx costwright-mcp with the COSTWRIGHT_API_KEY environment variable. Four tools are provided: costwright_check (static analysis), costwright_certify (signed certificate), costwright_verify (public certificate verification), and costwright_pubkey (retrieve public keys). The verify and pubkey tools require no API key.
Key features of Costwright Mcp
- Static worst‑case token‑budget analysis for LLM agent workflows.
- Reports certifiable, default‑dependent, non‑certifiable, and runaway units.
- Issues Ed25519‑signed budget certificates logged to a transparency log.
- Pure AST analysis – no code execution required.
- Compatible with LangGraph, CrewAI, and OpenAI Agents SDK.
- Public
verifyandpubkeytools for offline verification.
Use cases of Costwright Mcp
- Catch missing token caps before deploying an agent workflow.
- Identify
while True:runaway drivers in LLM agent code. - Provide a “budget type check” for agent workflow pipelines.
- Issue signed budget certificates for compliance or auditing.
FAQ from Costwright Mcp
What kind of repositories does Costwright Mcp analyze?
It analyzes Python repositories that use LangGraph, CrewAI, or the OpenAI Agents SDK, returning worst‑case token budgets for every workflow graph.
Does Costwright Mcp run the code it analyzes?
No. It performs pure static AST analysis – the code is never executed.
Is an API key required to use Costwright Mcp?
Yes, for the check and certify tools. The verify and pubkey tools are public and require no key.
Where does my source code go when using Costwright Mcp?
The check and certify tools upload a compressed archive of the repository to the hosted API. See the privacy policy at https://eleata.io/privacy/ for details.
What output does Costwright Mcp produce?
It returns a pass/fail status along with counts of units categorized as certifiable, default‑dependent, non‑certifiable, or runaway, and identifies which LLM calls have no token cap.
