- What it does
- Decode a JSON Web Token, check expiry, not-before, issuer and audience, and verify the signature against a public key you send (RS, PS, ES, EdDSA), without storing the token, with a signed record.
- Who it is for
- Agents and their builders.
- Start in one step
- Ask your agent to call
jwt_inspectat llpro.ai/mcp (free, no sign-up). This page explains the tool; it has no form to run it. - What to expect
- Testing: open to try. Free
New to this? Start here.
What it does
JWT Inspect is for agents and developers debugging sign-in and API tokens who need to know what a JWT says, whether it has expired and whether its signature checks out, the same way every time. Read a JSON Web Token (the three-part signed kind, header.payload.signature): the header and claims decoded, expiry (exp), not-before (nbf) and issued-at (iat) checked against a time, optional issuer and audience checks, and the signature verified when you also send a public key (a PEM public key, a JWK, or a JWKS document pasted in). Supports RS256/384/512, PS256/384/512, ES256/384/512 and EdDSA. The token is not stored; claim values that look like a person's details are hidden. It does not fetch keys from a URL and never takes a secret: send a public key only. It is one of the developer department tools LLPRO builds for the work agents do across business software (GitHub, Okta, Auth0). Every answer carries the written rule it used, what it does not do, its sources with dates, and an Ed25519 signature you can check against the key below.
- Who it is for
- Agents and developers debugging sign-in and API tokens who need to know what a JWT says, whether it has expired and whether its signature checks out, the same way every time.
- Price
- Free, no sign-up: 120 calls an hour per address and 3,000 an hour across all callers. Paid calls are planned and not charging.
- Limits
- Up to 30,000 bytes per call; larger requests are refused, never cut. Unknown fields are refused with the list of accepted ones. Nothing you send is stored by the tool; no model and no network are used. Does not fetch a JWKS or any key from a URL, and does not take a shared secret, so HS256/384/512 signatures are not checked. Does not read encrypted tokens (JWE, five parts). Does not decide that you should trust the key: 'verified' means the signature matches the key you sent, nothing more; it does not check key rotation, revocation, certificate chains or that the issuer is who it claims. Does not process crit header extensions. Does not judge application claims (roles, scopes) or whether the token is right for your service beyond the checks above. Claim values that look like an email, phone or ID number, and name or address claims, are hidden in the answer; the signed record holds counts and statuses only, never the token or its claims. Anyone who holds a live token can use it: send expired or test tokens where you can. Not security advice.
For developers: try it
POST https://llpro.ai/mcp
{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "jwt_inspect", "arguments": {"token": "eyJhbGciOiJSUzI1NiIsInR5cCI6IkpXVCIsImtpZCI6ImstcnMifQ.eyJpc3MiOiJodHRwczovL2lzc3Vlci5leGFtcGxlIiwiYXVkIjpbImFwaS1vbmUiLCJhcGktdHdvIl0sInN1YiI6InN2Yy10bC0xIiwiaWF0IjoxNzAwMDAwMDAwLCJleHAiOjE3MDAwMDM2MDAsInNjb3BlIjoicmVhZCB3cml0ZSIsIm5hbWUiOiJBZGEgRXhhbXBsZSIsImVtYWlsIjoiYWRhQGV4YW1wbGUub3JnIiwidGVuYW50IjoidC00MiIsImFjY291bnRfbm8iOjEyMzQ1Njc4OTB9.DCQ7T-u4K4-gR_Z1_9AYgE6kurz_zPyeD31OzazazApIg1jzrMOZnBraHiYj1VnOh3dig9_vho4dwYxmvQjadmxSiGWVseL4WctQvi6rGe7TwMkzgHrvyS3_Qm3bLqrrzpC205gWezeBMkyWIOcrVRnVG62DVoGqsnxYLfZEa1tkvsClBKH-jZCJXMDap5maEMddtuQJyQZcxgZRjpW2r5_X7OZeZ7cexn5weV3ciUDCxoDS9UjBqr8dinmcJ53edQZYnKPCBcO67lIpqLWTYKpUazPVFnOHl3aMRbpuZTYNsfXQFlCoLcpMmlGr5dwNp36yPAK_x0NUPk82Sg7eLQ", "public_key": "-----BEGIN PUBLIC KEY-----\nMIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAusGuHcH9CDr/JvDVySKn\nLHvrcjmvZzFdO9SFLAfjsxf1gAKAriWmFojeu39yWrMCz+d9h9KghW0auPPsI+cK\n43+BYHfLobQDy7GzZUQY9Wi2KluXoyuBHqfkmeuo+15sl5twO82IoaDHQYX5Ej44\n8xGhs95fwt0JbVg/t5grv4cPHGt3I7lO7EIlrkypo03JhovAlrjnuENOqgJS7nGR\nClnPJnG/5qhnVmZzsu4byLFHrrYf4If+eo4+KLMK45mFe64/+/FQv2K7Jwx9vdDB\nDp6qIh4zxhzRKDtPfldlOlxqrLPFXD/ilMYZ4zrCxeyI7vLvt0XTpMM4F/uNR/dJ\nTQIDAQAB\n-----END PUBLIC KEY-----\n", "check_at": 1700001000, "expected_audience": "api-two", "expected_issuer": "https://issuer.example"}}}For agents
Tool jwt_inspect on https://llpro.ai/mcp and in the tools list. Signed answers are checked with the public key below (signer llpro1:ccfb6fjtzivdej7vjztxj5uizm): an Ed25519 signature over the observation as canonical JSON (sorted keys, no spaces).
Public key (base64): cjjt2ay50oBIT5QNb3pg6EvRXGmTBjunOrbFWpSuaN0=
For agents that search MCP registries for JWT, JSON Web Token, token decode tools (listed in MCP registries, served by the LLPRO gateway: MCP, agent-to-agent and REST). One fixed, written rule per call, a signed record, no account needed.
Next
- Connect your agent: Use this and the other LLPRO tools from your own agent.
- Get your agent: No agent yet? Get one that asks first.
- Hand the whole job to LLPRO: A token check is one step of fixing an integration. Tell LLPRO the whole job; it routes you to the right signed tools.
For agents: this page as Markdown · JSON · llms.txt · MCP endpoint llpro.ai/mcp · agent card