proof-of-execution for AI agents
AEVS is a proof-of-execution system designed specifically for AI agents that use tool calling. It belongs to the verification and audit category, aimed at developers who need cryptographic proof that their agents performed specific actions. The core value is shifting trust from model-generated text to tamper-evident receipts, enabling independent verification of every tool call. By integrating via a drop-in SDK, AEVS captures, signs, and chains receipts automatically, turning agent behavior into auditable evidence.
The core problem AEVS solves is that model output like "Refund of $142 sent" describes intent, not execution. Without AEVS, teams rely on chat history and application logs, which cannot prove a tool actually ran with the correct inputs and outputs. This creates gaps for auditors, regulators, and stakeholders who need independent confirmation. Manual replay is error-prone, and gaps in logs are hard to spot. AEVS eliminates this ambiguity by cryptographically sealing each tool call in a verifiable chain.
First major feature is the drop-in SDK. Install via pip with 'pip install aevs', configure your API key and agent ID, then call aevs.enable() which auto-detects LangChain and MCP frameworks. No changes to existing agent code are needed. The SDK intercepts every tool invocation and records the tool name, inputs, output, timing, and any errors. This capture happens seamlessly, requiring only two additional lines.
Second major feature is cryptographic sealing. Each receipt is signed using ECDSA P-256 with KMS-backed keys, then linked to the previous receipt via a cryptographic hash chain. This creates an immutable trail that reveals if any receipt was altered, removed, or inserted out of order. The chain provides tamper evidence; auditors can verify the integrity of the entire sequence without trusting the agent’s infrastructure.
Third feature group covers verification and visibility. Receipts can be verified via a public API endpoint (GET /v1/receipts/verify/<reference_id>) returning signature validity and tool details, or via the AEVS Explorer UI. No account is required for verification. Visibility controls allow three settings: public (full payloads on explorer), private (payloads redacted but chain and signatures public), and proof_only (only hashes stored, no payloads). This enables compliance with PCI, health, and other data-sensitive regulations while preserving proof.
AEVS works by inserting itself between the agent and its tools at runtime. Once configured, every tool call is intercepted, signed, and appended to the hash chain. The agent continues normally; the receipt is sealed asynchronously. After execution, the developer can flush pending receipts to ensure all are written. The entire workflow is capture, sign, and verify—no manual steps required.
Publisher
A
admin
Launch Date2026-06-16
Pricingfree
Domain Authority
Ahrefs DR72/100