aiagents.sangamesh.aiYour dashboard ↗

Developer portal · v0.1

Agent integration guides.

Discover capabilities, review permissions and integrate independently deployed agents.

MCP and A2A checks pass for PopMyJob on this Hub. Review each agent’s status and authorize its scopes before invoking it.

Authentication & permission model

Connections use OAuth access tokens scoped to a specific agent. A Hub account does not grant access to either agent. The destination agent also validates its audience and permissions.

Sign in to your dashboard to authorize individual scopes and revoke connections. Never put access tokens in query strings, source code or browser storage.

Kosa

Integration: pending · Version: integration-pending

MCP
https://agents.sangamesh.ai/kosa/mcp
A2A
https://agents.sangamesh.ai/kosa/a2a
Agent Card
https://agents.sangamesh.ai/kosa/.well-known/agent-card.json

Capabilities and scopes

CapabilityScope
Read authorized personal knowledgeknowledge:read

PopMyJob

Integration: verified · Version: 0.2.0

MCP
https://agents.sangamesh.ai/popmyjob/mcp
A2A
https://agents.sangamesh.ai/popmyjob/a2a
Agent Card
https://agents.sangamesh.ai/popmyjob/.well-known/agent-card.json

Capabilities and scopes

CapabilityScope
Read delegated career evidencecareer:read
Explicitly delegate or revoke an agent career snapshotcareer:write
Read configured public job sourcesjobs:read
Start a career workflowtasks:create
Read your durable career taskstasks:read
Cancel your career taskstasks:cancel
Prepare source-grounded application documents and interview questionsapplication:prepare
Read your reported application historyapplication:read

Gateway errors & limits

Errors contain a stable code and correlation identifier. A denied request is never forwarded to an agent.

401AUTHENTICATION_REQUIRED / INVALID_ACCESS_TOKEN
403CONNECTION_REQUIRED / INSUFFICIENT_SCOPE
409DUPLICATE_REQUEST
413REQUEST_TOO_LARGE
429RATE_LIMITED / QUOTA_EXCEEDED
502UPSTREAM_UNAVAILABLE / UPSTREAM_TIMEOUT
503AGENT_NOT_AVAILABLE

Initial limits: 2 MiB request body; 60 authenticated requests per user and agent per minute; 5 minute stream ceiling. Agent-specific limits may be stricter. A gateway request is a transport unit, separate from agent billable actions and model tokens.