# @agentguard-run/spend ## Functional-Use Disclaimer DISCLAIMER: All terminology and labels used in AgentGuard’s materials are descriptive of software functionality only, not legal definitions or guarantees of compliance. For example, the terms “receipt,” “audit log,” “evidence,” “audit trail,” and “attestation” refer solely to cryptographically-signed records produced by AgentGuard’s software. These terms do not mean that any record is legally binding evidence, certified by any authority, or equivalent to records maintained by banks, auditors, or courts. Similarly, references to “signed,” “verified,” or “attested” pertain to digital signature processes, not to notarization or governmental attestation. Any use of the word “compliance,” “compliant,” “outcome,” or “settlement” describes the software’s features in a non-regulatory sense. In particular, “capability tier” is merely an internal category of agent permission levels, and does not correspond to any government or industry regulatory classification. Likewise, any statement that an operation “settles” or an outcome is “settled” refers only to the software’s final state, not to any legal settlement. No AgentGuard tool or document should be interpreted as providing legal, financial, or regulatory advice, or as a certified compliance certificate. AgentGuard is not a law firm or auditor and does not guarantee adherence to laws or standards. All use of AgentGuard software is at the user’s risk, and users must ensure independently that their use of AI agents complies with all applicable laws and regulations.* (CONTINUED IN NEXT PARAGRAPH) Throughout this documentation, package metadata, CLI banners, and related materials, any word drawn from compliance or legal contexts (such as “audit,” “evidence,” “settlement,” “credit,” “debt,” “market,” “liquidity,” “maturity,” or similar) is used only in a functional, descriptive sense. For instance, “audit log” simply means a sequence of recorded events; it does not imply an official financial or regulatory audit. The phrase “satisfies” is used to indicate that one software condition meets a programmed check (e.g. satisfying a budget cap), and is not intended to suggest that any legal requirement has been fulfilled. Under no circumstances should AgentGuard’s terminology be taken to indicate that our software is executing regulated financial or legal processes.* No term in our code, documentation, or marketing is meant to designate or characterize any AgentGuard feature as a service subject to banking, securities, derivatives, insurance, real estate, or other regulated activities. For example, “trade,” “trading,” “liquidity,” “maturity,” and similar terms (if ever used) describe algorithmic budget or token flows, not financial transactions. AgentGuard does not act as a broker-dealer, clearinghouse, insurance issuer, or government agency. References to “market” or “liquidation” are metaphorical descriptions of how the software handles tokens or budgets. No AgentGuard action should be construed as offering financial services or operating a regulated marketplace.* Likewise, words like “offer,” “obligation,” “credit,” “debt,” “payment,” or “settle” in our text refer only to hypothetical or illustrative actions of a user’s AI agent or simulated transaction flows. They do not mean that AgentGuard is engaging in any money-handling, offering financial instruments, or guaranteeing any settlement of obligations. If AgentGuard generates an invoice or cost-of-service figure, it is purely illustrative of resource usage, not a binding financial bill. The term “maker/taker” (if used in examples) is purely a naming convention and has no relation to exchange regulation. The software’s “payment execution” tier is only a capability checkbox, not an actual payment processor. AgentGuard does not transmit money or credit.* Finally, any references to “certify,” “verify,” or “attest” should be read in context of cryptography and computing. For example, when we say a receipt is “signed” or “verified,” we mean via public-key cryptography. AgentGuard does not claim that receipts are admissible legal evidence by default. Users remain responsible for any legal implications of presenting AgentGuard logs or receipts in regulatory audits or court. This disclaimer is intended to ensure AgentGuard’s terms (in this README, license, package descriptions, CLI banner, customer agreements, and /llms.txt) make clear we provide a technical audit tool only. The functionality is delivered “as-is” with no implied regulatory endorsement. > **AgentGuard proves what your AI agent attempted, who authorized it, what it cost, and whether it succeeded.** > > Local-runtime spend caps, capability-gated model routing, and Ed25519-signed receipts for AI agents. > Also available in: [Español (LATAM)](README.es-419.md) · [Português (BR)](README.pt-BR.md) Every policy decision runs inside your process. Prompts, completions, provider API keys, signing keys, policies, and cost overrides never go to AgentGuard infrastructure. OpenRouter calls go directly from your runtime to `openrouter.ai` with your key. ## Quickstart in 90 seconds ```bash npm install @agentguard-run/spend agentguard auth openrouter agentguard wizard ``` The wizard writes: - `~/.agentguard/policy.yaml` - `~/.agentguard/quickstart.ts` - `~/.agentguard/quickstart.py` It also prints the snippet to paste into your app: ```ts const response = await guardedClient.chat.completions.create({ model: 'openai/gpt-4o-mini', messages: [{ role: 'user', content: 'Run the governed task.' }], }); ``` Then run your agent. AgentGuard decides locally before any provider call starts, signs the receipt, and applies allow, downgrade, shadow, or block. By default the wizard wires a file-backed `NdjsonDecisionLogStore` and a locally generated Ed25519 signing key, so every decision is signed and appended to `~/.agentguard/my-tenant/decisions.ndjson` with no extra setup. ### See it: run `agentguard serve` After your agent runs (or after `agentguard demo`), open the local dashboard to SEE governed spend, the dollars saved by policy, per-agent and per-provider breakdowns, and a tamper-evidence badge with click-to-verify receipts: ```bash agentguard serve # → dashboard at http://localhost:8787 ``` The dashboard reads only your local `~/.agentguard/*/decisions.ndjson`. No outbound network calls. ## Workflow-level caps Use `agentguard.workflow()` when one agent run spans many outcomes and needs one shared budget envelope. The SDK signs each step as receipt schema v2, writes checkpoints, and validates the prior chain before resume. ```ts import { agentguard } from '@agentguard-run/spend'; await agentguard.workflow({ name: 'migrate_legacy_payments', budget_cap_usd: 500, duration_cap_hours: 72, checkpoint_every_outcomes: 50, resume_if_exists: true, }, async (workflow) => { for (const record of records) { await workflow.outcome('migrate_payment', () => migratePayment(record)); } }); ``` Share the public replay URL at `https://agentguard.run/verify/workflow/`. ## Need help configuring? Run `agentguard advisor` `agentguard advisor` uses your OpenRouter key, or any OpenAI-compatible endpoint you pass with `--base-url`, to run a local setup dialogue in your terminal. AgentGuard infrastructure never sees the prompts, completions, keys, policy details, or session log. ```bash agentguard advisor ``` Advisor writes `~/.agentguard/policy.yaml`, a language-aware quickstart file, projected savings math, and a local JSONL session log under `~/.agentguard/advisor-sessions/`. ## Governance Posture Advisor asks for one operating-style input and uses it to shape the generated policy. - `velocity`: high-ship software and AI teams. Starts in `shadow`, uses permissive capability tiers, and downgrades aggressively to cheaper models. - `standard`: most SaaS, e-commerce, real estate, agencies, local services, and startups. Starts in `enforce`, uses balanced capability tiers, and keeps 90 days of audit retention. - `compliance`: law, healthcare, dental, accounting, SOX, fintech, and regulated workflows. Starts in `canary`, requires stricter capability tiers, blocks regulated overflow instead of downgrading, and keeps 7 years of audit retention. Override the suggestion when you already know how the team operates: ```bash agentguard advisor --posture velocity agentguard advisor --posture compliance ``` `custom` posture is reserved for the Solo tier Outcome Builder. ## Why OpenRouter? One OpenRouter key gives your team access to hundreds of models across many providers. Your CFO sees one invoice. AgentGuard enforces who uses what, which task tiers can reach which models, and what each call can spend. The OpenRouter key can live in `OPENROUTER_API_KEY` or `~/.agentguard/openrouter-key` with mode `600`. Sync pricing when you want local cost math refreshed: ```bash agentguard models --sync-pricing agentguard models --task payment-approval agentguard models --search gpt-4o --json ``` Pricing overrides are stored locally in `~/.agentguard/cost-overrides.json`. ## Verify any receipt Share https://agentguard.run/verify with an auditor or reviewer. Paste a receipt and public key in the browser to verify the Ed25519 signature, entry hash, and chain link. The receipt never leaves the page. CLI verification is still local: ```bash agentguard demo agentguard verify --trace latest ``` ## Task templates `agentguard wizard` ships templates for: - `risk-review`: read-only review with a $0.50 per-call cap - `payment-approval`: payment initiation review with a $5.00 per-call cap - `chargeback-evidence`: evidence assembly with a $1.00 per-call cap - `agent-support`: data-write support workflow with a $0.25 per-call cap - `code-scan`: long-context read-only scan with a $0.10 per-call cap Each template sets recommended OpenRouter model assignments, capability tier, fallback model, caps, and system instructions. ## Provider bindings TypeScript includes native OpenAI, Anthropic, and Bedrock bindings. Each wraps a direct provider client in one line, with per-provider cost tables and response-shape handling. Enforcement semantics are identical across all three (and the OpenRouter path): the same policy blocks, downgrades, shadows, or allows a call before any bytes leave the process. Streaming usage is settled from provider usage events when available, with local token-estimator fallback when usage is missing. Settlement entries are signed into the same hash chain as enforcement decisions. Wrap a direct **OpenAI** client: ```ts import OpenAI from 'openai'; import { withSpendGuard } from '@agentguard-run/spend'; const openai = withSpendGuard(new OpenAI(), { policy, scope: { tenantId: 'acme' } }); await openai.chat.completions.create({ model: 'gpt-5', messages }); ``` Wrap a direct **Anthropic** client: ```ts import Anthropic from '@anthropic-ai/sdk'; import { withSpendGuardAnthropic } from '@agentguard-run/spend'; const anthropic = withSpendGuardAnthropic(new Anthropic(), { policy, scope: { tenantId: 'acme' } }); await anthropic.messages.create({ model: 'claude-sonnet-4-6', max_tokens: 1024, messages }); ``` Wrap a direct **Bedrock** runtime client: ```ts import { BedrockRuntimeClient } from '@aws-sdk/client-bedrock-runtime'; import { withSpendGuardBedrock } from '@agentguard-run/spend'; const bedrock = withSpendGuardBedrock(new BedrockRuntimeClient({ region: 'us-east-1' }), { policy, scope: { tenantId: 'acme' } }); await bedrock.send(new InvokeModelCommand({ modelId: 'anthropic.claude-sonnet-4-v1:0', body })); ``` Only `InvokeModel` / `InvokeModelWithResponseStream` commands are gated; other Bedrock commands pass through untouched. Python includes OpenAI, Anthropic, Bedrock, LangChain, CrewAI, and LlamaIndex integration helpers. ## Framework integrations Drop-in adapters translate each framework's own hook/middleware surface into the same shared enforcement core (`guard.decide()` + the Ed25519-signed receipt chain). They never proxy prompts, completions, or provider keys — only redacted request shape reaches a receipt. One line each: ```ts // Vercel AI SDK — wrapLanguageModel middleware (runs guard.decide before generate/stream) import { wrapLanguageModel } from 'ai'; import { agentguardAiSdkMiddleware } from '@agentguard-run/spend/frameworks/vercel-ai'; const governed = wrapLanguageModel({ model, middleware: agentguardAiSdkMiddleware({ policy, scope }) }); // OpenAI SDK / OpenAI Agents SDK — wrap the client's chat.completions.create import OpenAI from 'openai'; import { withSpendGuardOpenAI } from '@agentguard-run/spend/frameworks/openai'; const client = withSpendGuardOpenAI(new OpenAI(), { policy, scope }); // LangChain (JS) — BaseCallbackHandler (handleChatModelStart preflights, handleLLMEnd settles) import { ChatOpenAI } from '@langchain/openai'; import { createLangChainHandler } from '@agentguard-run/spend/frameworks/langchain'; const llm = new ChatOpenAI({ callbacks: [createLangChainHandler({ policy, scope })] }); // OpenRouter — wrap an OpenAI-compatible client pointed at openrouter.ai import { withSpendGuardOpenRouter } from '@agentguard-run/spend/frameworks/openrouter'; const or = withSpendGuardOpenRouter(new OpenAI({ baseURL: 'https://openrouter.ai/api/v1' }), { policy, scope }); // Claude Code — PreToolUse/PostToolUse hooks import { createClaudeCodeHooks } from '@agentguard-run/spend/frameworks/claude-code'; const hooks = createClaudeCodeHooks({ policy, scope }); // Hermes — pre-tool-call / approval / post-tool-call plugin hooks import { createHermesPlugin } from '@agentguard-run/spend/frameworks/hermes'; export default createHermesPlugin({ policy, scope }); ``` ## Spend-aware model router (removes vendor risk) `createRouter(...)` composes the bindings, the enforcement engine, and the signed receipt chain into one call: it picks a candidate model across providers by policy, enforces the spend cap with the **same** enforcement path (block / downgrade / allow), executes it, and **fails over to the next provider** on a provider error/timeout or a cap breach on the chosen model. Every route is recorded in a signed, hash-chained routing receipt (candidates considered, winner, reason, resolved cost, fallbacks) that verifies with the existing `verifyChain`. No prompt/completion content or keys ever enter the receipt. Selection policies: `cost` (cheapest capable first), `priority` (explicit failover order), or `capability` (filter by a capability tag, then cheapest-first). ```ts import OpenAI from 'openai'; import Anthropic from '@anthropic-ai/sdk'; import { createRouter } from '@agentguard-run/spend'; const router = createRouter({ candidates: [ { provider: 'anthropic', model: 'claude-sonnet-4-6', client: new Anthropic() }, { provider: 'openai', model: 'gpt-5-mini', client: new OpenAI() }, ], policy: { mode: 'cost' }, // cheapest capable first; failover in order on error/cap spendPolicy: policy, // your existing SpendPolicy — enforcement is REUSED, not forked signingKeys, // same Ed25519 keys as the rest of the SDK logStore, // routing receipts hash-chain onto the same signed log }); const { response, chosen, reason, resolvedCents, fallbacks, routingReceipt } = await router.route({ messages, params: { max_tokens: 1024 } }); // chosen -> { provider, model }, reason -> 'cost' | 'priority' | 'capability' | 'failover' | 'downgrade' ``` Try it offline (mock providers, real signatures): `agentguard route-demo`. ## Sample receipts for outreach Generate a realistic, signed sample receipt for a named flow plus a self-contained, self-verifying HTML page — "exactly what your counterparty would get": ```bash agentguard sample --flow x402-payment # or: refund | agent-purchase ``` This writes `receipt.json` and `verify.html`. The HTML re-derives the entry hash from canonical JSON (SHA-256) and verifies the Ed25519 signature in-browser via WebCrypto, with NO network calls; where a browser lacks WebCrypto Ed25519 it still checks the hash chain and links to agentguard.run/verify with the receipt prefilled. The signature is real; only the flow parameters are illustrative. ## External timestamp anchoring (opt-in) Signed receipts prove internal order and integrity, but the signatures are produced in your own environment. Anchoring the chain head's SHA-256 to a public timestamp calendar (OpenTimestamps → Bitcoin) adds an EXTERNAL clock, so "this chain existed no later than