Gvnr

Substrate primitives for AI agents — spend caps, rate limits, idempotency, post-call reconciliation, human approval bridges. One MCP endpoint, one credit pool, no infrastructure to deploy.

Listed on the Official MCP Registry as dev.gvnr/gvnr.

No deployment. No proxy. No self-hosting.

The problem

Agents cost 10–12x more than estimated in production. System prompts, retry loops, and tool calls multiply fast. A runaway agent can generate a $47,000 bill in 11 days. The common fix — self-hosting LiteLLM — requires running infrastructure most developers won't set up.

Gvnr is the hosted alternative: an external authority your agent checks before spending.

How it works

1. Your agent calls budget_clear (MCP tool or REST) before each LLM request

2. The governor checks your account credit balance and the agent's spend envelope

3. It returns { approved: true } or { approved: false, reason: "..." }

4. If denied, your agent skips the call

The envelope is configured by you (per-agent daily or session cap). The credit balance is topped up via USDC on Base.

Quick start

1. Provision an account

curl -X POST https://gvnr.dev/v1/account
# { "api_key": "bg_...", "account_id": "..." }

2. Top up credits

Open the payment page for your chosen pack, pass your API key as a query param:

https://gvnr.dev/pay/starter?api_key=bg_YOUR_KEY

Send USDC on Base to the address shown, paste your tx hash — credits are added after on-chain verification.

Or, if you prefer the programmatic path — POST the tx hash directly:

curl -X POST \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"tx_hash":"0x..."}' \
  https://gvnr.dev/v1/account/topup-verify/starter

3. Set a spend envelope for your agent

curl -X PUT \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"my-agent","limit_usd":5,"window":"daily"}' \
  https://gvnr.dev/v1/budget/envelope
# { "success": true, "agent_id": "my-agent", "limit_usd": 5, "window": "daily" }

4. Set a rate envelope per (agent, provider, model)

curl -X PUT \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"my-agent","provider":"anthropic","model":"claude-sonnet-4-6","requests_per_minute":30}' \
  https://gvnr.dev/v1/rate/envelope
# { "success": true, ... }

5. Before each LLM request: budget_clear then rate_check

curl -X POST \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"my-agent","model":"claude-sonnet-4-6","estimated_tokens":2000}' \
  https://gvnr.dev/v1/budget/clear
# { "approved": true, "remaining_usd": 4.994 }

curl -X POST \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"my-agent","provider":"anthropic","model":"claude-sonnet-4-6"}' \
  https://gvnr.dev/v1/rate/check
# { "allowed": true, "requests_remaining_this_minute": 29 }

6. (optional) Dedupe retries with idempotency_check

curl -X POST \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"key":"job-abc-123","ttl_seconds":3600}' \
  https://gvnr.dev/v1/idempotency/check
# First call:  { "is_first_call": true,  "ttl_remaining_seconds": 3600 }
# Replay:      { "is_first_call": false, "ttl_remaining_seconds": 3598 }

7. After the LLM responds, reconcile against actual usage

curl -X POST \
  -H "Authorization: Bearer bg_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"my-agent","actual_input_tokens":1800,"actual_output_tokens":2400}' \
  https://gvnr.dev/v1/budget/reconcile
# { "ok": true, "drift_usd": 0.003, "remaining_usd": 4.991, "balance_usd": 9.991 }

reconcile is optional but keeps the envelope honest — Anthropic, OpenAI, and Gemini all return usage fields with the actual token counts; pass those in.

MCP setup

Add to Claude Desktop or any MCP-compatible client:

https://gvnr.dev/mcp?api_key=bg_YOUR_KEY

Claude Code

claude mcp add gvnr --transport http \
  "https://gvnr.dev/mcp?api_key=bg_YOUR_KEY"

MCP tools

REST API

All endpoints (except POST /v1/account) require Authorization: Bearer bg_YOUR_KEY.

Account

Budget

Clearance response

{ "approved": true, "remaining_usd": 4.994 }

{ "approved": false, "remaining_usd": 0, "reason": "envelope_exceeded" }

Denial reasons: no_credits · no_envelope · envelope_exceeded

Credit packs

Top up at GET /pay/:pack?api_key=bg_YOUR_KEY. Send USDC on Base mainnet — credits added after on-chain verification.

Envelope windows

• daily — resets at UTC midnight each day

• session — never resets (use for one-shot tasks)

Supported models

Model pricing is a static lookup on the hot path — no external calls. Includes claude-opus-4-7, claude-sonnet-4-6, claude-haiku-4-5, gpt-4o, gpt-4o-mini, and others. Embedding models (text-embedding-3-small/large, gemini-embedding-001/2) are billed input-only — pass input tokens to budget_clear for those. Unknown models fall back to a conservative default.

Network

Current deployment: Base mainnet.

License

MIT — see LICENSE.

The canonical hosted service is at https://gvnr.dev. Self-hosted instances are unaffiliated.