get_project_truthREADRead current verified and likely Truth claims backed by the latest completed scan.
↗HARIKOS exposes a small, agent-neutral tool contract for current Truth, relevant Memory, recent Changes, Context Packs, assumption checks, and structured write-back.
{
"mcpServers": {
"harikos": {
"url": "https://harikos-ai.vercel.app/api/mcp/PROJECT_ID",
"headers": {
"Authorization": "Bearer TOKEN_SHOWN_ONCE"
}
}
}
}This marketing-only terminal demonstrates the shape of HARIKOS output. It is clearly illustrative and does not pretend to be a connected production CLI.
Authentication
Supabase Auth
VERIFIED · 99%
Evidence
middleware.ts:18
lib/supabase/server.ts:7
Previous: Clerk · SUPERSEDED
Each tool has a narrow job and verifies authorization before returning project data.
get_project_truthREADRead current verified and likely Truth claims backed by the latest completed scan.
↗search_project_memoryREADSearch recorded project Memory by content or type.
↗get_recent_changesREADRead the ten most recent repository changes in the verified snapshot.
↗get_context_packWRITECreate and persist a task-specific Context Pack. This uses the project's Context quota.
↗record_memoryWRITEPersist a project Memory entry. This uses the project's Memory-write quota.
↗record_outcomeWRITERecord a session outcome and durable Memory for completed agent work.
↗check_assumptionREADCheck a statement against current project Truth before acting.
↗begin_agent_sessionSESSIONOpen an auditable project-scoped agent session, optionally with its task.
↗end_agent_sessionSESSIONClose a recorded agent session as completed, failed, or abandoned.
↗The agent receives labeled project knowledge, not an opaque blob of retrieved text.
POST /api/mcp/:projectId
Authorization: Bearer ...
{
"method": "tools/call",
"params": {
"name": "check_assumption",
"arguments": {
"statement": "Auth uses Clerk"
}
}
}{
"verdict": "contradicted",
"currentTruth": "Supabase Auth",
"status": "verified",
"evidence": [
"middleware.ts:18-34"
]
}Connect a repository and give every coding agent one current, evidence-backed understanding.