> How do you let an agent use a tool that needs credentials without giving the credentials to the agent?
I took a different approach [0].
1. Register a secret and get back an opaque identifier (out-of-band, human action; can extend and add an API for listing by endpoint)
2. Generate a JavaScript API call that includes the opaque identifier as header (for example)
3. Send the JavaScript to a server that runs a JS interpreter
4. Pre-process the inbound JS and replace the opaque identifier with the actual token
5. Run the JS to invoke the API, make transforms, process the result, etc.
Effectively using agent generated JS script to make the actual API call where the agent only sees the opaque ID of the secret.
If the server is compromised, it still receives the real key on each call. If the vault issues a long lived key, e.g. GitHub PAT, one call will expose it forever. Does the vault issue short lived keys or the stored one?
The audit evidence is hash chained and we allow for it to be exported off platform. Currently s3 bucket is supported. Additionally, we allow for normal csv-based exports. Would be interested to know what centralized log collection services you use.
[0] https://github.com/CharlieDigital/runjs
If the server is compromised, it still receives the real key on each call. If the vault issues a long lived key, e.g. GitHub PAT, one call will expose it forever. Does the vault issue short lived keys or the stored one?
This is great, we are working on something similar at 1Claw but I like your approach too. Want to trade notes sometime?
How would someone send the audit evidence to their organization's centralized log collection service(s)?
The audit evidence is hash chained and we allow for it to be exported off platform. Currently s3 bucket is supported. Additionally, we allow for normal csv-based exports. Would be interested to know what centralized log collection services you use.
Other MCP gateway platforms seem to send extensive records to SIEM systems https://docs.domesystems.ai/operate
Good point. What concerns do you see with this approach? And how would you handle sending this kind of activity to SIEM?
[flagged]
[dead]
[flagged]
[dead]
[dead]