Inficy

What is Inficy?

Inficy is an evidence platform for teams that run AI agents in production. It captures each agent run from runtimes such as Codex CLI, Claude Code, and Hermes, then turns it into a tamper-evident record showing steps, tool calls, failures, handoffs, and human approvals. Eligible runs can be sent for independent certification.

Top Features:

• Runtime capture: records agent steps and tool calls without proxying your model provider traffic.
• Tamper-evident bundles: each completed run is sealed so later edits become detectable.
• Fleet analytics: dashboards show time spent, failures, retries, and human interventions across agents.

Use Cases:

• Audit trails: keep verifiable records of agent actions for compliance and internal reviews.
• Failure debugging: trace errors, retries, and recoveries in an ordered execution timeline.
• Payment oversight: see requested, approved, failed, and refunded payment states exactly as observed.

Who Can Use Inficy?

• AI engineering teams: inspect how coding and research agents behave across many runs.
• Compliance officers: collect exportable evidence of agent activity and human approvals for audits.
• Fintech operators: track agent-driven payments and approvals with a verifiable record.

Pricing

• Free ($0 per month): one workspace, unlimited agents, timelines, and one certification credit.
• Starter ($9 per month): longer retention, included certification credits, and email support for small fleets.
• Growth ($29 per month): higher ingestion limits, 365 days of retention, and priority support.

Pros and Cons

Pros:

• Free recording: activity capture and timelines are included on every plan.
• No traffic proxy: model provider keys and requests stay inside your own runtime.
• Honest data: failures stay visible and unobserved activity is never invented.

Cons:

• Limited runtimes: only supported connectors such as Codex CLI and Claude Code work today.
• Certification costs: extra credits cost between four and five dollars per certified run.
• Narrow scope: certification confirms integrity but does not prove agents acted correctly.

FAQs:

1) Does it proxy my model traffic?

No, your OpenAI, Anthropic, or xAI keys and requests stay inside your own runtime.

2) Which agents can it record?

Supported runtimes include Codex CLI, Claude Code, and Hermes, plus a ChatGPT connector for sessions.

3) What does certification prove?

It attests the stored record is unchanged, not that the agent was correct or compliant.

4) Is there a free plan?

Yes, the Free plan includes recording, timelines, unlimited agents, and one certification credit.

5) Can I export the records?

Yes, you can export single executions or filtered sets as structured evidence bundles.