Praxel.
Runtime control plane

Govern every agent.
Signed by default.

Praxel is the vendor-neutral runtime control plane for AI agents — an open-source SDK that wraps Anthropic, OpenAI, or a custom adapter in three lines, then adds cryptographically signed audit logs, policy enforcement, and signed compliance evidence (SOC 2 / HIPAA / EU AI Act report generation shipping in v1.0).

before · agent.pyno audit trail
from anthropic import Anthropic

client = Anthropic()
resp = client.messages.create(
  model="claude-sonnet-4-5",
  tools=tools,
)
after · praxel wrappedpolicy active
from praxis import Praxis

client = Anthropic()
agent = Praxis.wrap(client)
# → allow · signed:ed25519:b1e4…c09a
Seeking design partners across fintech, healthcare, and legal AI
Architecture
Vendor-neutral by structure
Cryptography
Ed25519 signed events
Policy evaluation
Sub-50ms
License
MIT open source
How it works

Where production meets proof

Every model call routes through four deterministic stops — intercept, decide, execute, seal — with sub-50ms policy evaluation.

Praxis.wrap( Anthropic(), 'finops' )import
wrap openai.ChatCompletionwrap anthropic.Messageswrap custom.adapter
Three providers supported today (Anthropic, OpenAI, custom adapter). LangChain, CrewAI, OpenAI Agents SDK, and MCP support shipping in v1.1.

Three-line intercept

Wrap any client at import time. No agent rewrite, no proxy, no new runtime.

Policy decisionallow
rule · read_only.safeallow
rule · amount > $1,000approve
rule · pii.contains(ssn)deny
rule · finops.wire.maxdeny
4 rules matched · sub-50msJSON policies

Real-time policy

Allow, deny, or escalate to a human — per call, per tool, per cost envelope.

signed ledgered25519
{
  "event": "evt_01JKR4V2Q8",
  "tool":  "send_ach",
  "action": "deny",
  "proof": "soc2.cc6.1",
  "sig":  "ed25519:b1e4c09a…"
}
Ed25519 signed, individually verifiable. Merkle chain on the v1.1 roadmap · exported to S3

Sealed evidence

Signed, individually verifiable, exported. Your auditor queries the ledger directly.

The stakes

Testing is a snapshot.
Production is a stream.

Three scenarios that should keep your CISO awake

Composite scenarios drawn from publicly reported AI incidents (2024–2026). Praxel enforcement would have logged, blocked, or escalated each.

Incident · 02:14 UTCfinops
$50,000
wire authorized by agent

A scheduling agent chained lookup_vendor and send_ach and wired funds to a spoofed payee. No human. No audit trail that survived rotation.

ExposureSOX 404 · Reg E
Disclosure · T+11 dayshealthcare
14,200
PHI records leaked

A Claude triage agent echoed patient records into a third-party summarizer. Screenshots surfaced in a training set. HHS opened a breach investigation in two weeks.

ExposureHIPAA · HHS OCR
Runaway · 38 minplatform
204/min
stripe calls in retry loop

A three-agent workflow hit a retry loop: planner → executor → search. By the time rate limits tripped, the team had burned $11k and a customer circuit breaker.

ExposureSOC 2 CC7.2

86% of production AI agents go live without formal security review

Source: Pragatix AI Agent Security Report, 2026

€35M or 7% EU AI Act max penalty for prohibited practices · enforcement August 2026

Source: EU AI Act Article 99

35% of executives report they cannot stop a rogue agent in real time

Source: Pragatix AI Agent Security Report, 2026

Platform

Fast, vendor-neutral, governed

Praxel is built for operators shipping agents into regulated environments. Speed and proof are not mutually exclusive.

Sub-50ms policy evaluation

Synchronous policy evaluation runs in-process. No sidecar, no proxy hop, no production-latency tax.

Policy bundles
Slack approvals
Redaction engine

Workflow governance

Distribute signed policies across fleets. Approvals route to Slack, PagerDuty, or webhooks — with sub-second escalation.

Agents
12 active · 1 sandbox
Policies
4 deployed · 1 pending
Approvals
2 queued
Decisions · last 24h1,247 signed
allow 1,180approve 47deny 20

Sample dashboard view

Unified control plane

One dashboard across every provider, every tenant, every region. Filter by model, agent, or rule. Export to your SIEM in a click.

Praxel · your runtime layer

Self-host or SaaS

Self-host SDK and audit logs in your environment. KMS integration shipping with Enterprise tier.

Why Praxel

The referee can't be on a team

OpenAI now owns Promptfoo. Microsoft governs Azure. Google will ship its own. The companies building AI agents cannot also be the neutral layer that governs them. Praxel is structurally neutral by design.

Vendor-neutral by design

MIT open source · vendor-neutral by structure. No frontier lab owns the roadmap.

Cryptographic evidence

Ed25519 signed events, individually verifiable, cryptographically timestamped. Merkle chain on the v1.1 roadmap. Tamper-evident by construction.

Runtime, not CI

Every call, every tool, every retry. Not a quarterly snapshot run in your eval pipeline.

Policy-as-code

Policy-as-code, JSON-configured (OPA bundle support on the v1.1 roadmap). Version-controlled, CI-tested, distributed with a signed git push.

Three providers supported today

Anthropic, OpenAI, custom adapter. LangChain, CrewAI, OpenAI Agents SDK, and MCP support shipping in v1.1. No rip-and-replace.

Audit-ready evidence

SOC 2, HIPAA, EU AI Act, OWASP — mapped, auto-generated, exported to your auditor's dashboard.

Integrations

Wrap anything. Govern everything.

One adapter pattern for any provider. Anthropic, OpenAI, and custom adapters supported today. LangChain, CrewAI, OpenAI Agents SDK, and MCP shipping in v1.1.

Compare

Why Praxel beats locked-in

Promptfoo tests agents before they ship. Microsoft governs agents in Azure. We govern agents in production, across every vendor.

Praxel
  • Vendor-neutral — Anthropic, OpenAI, custom adapter; no parent lab
  • Runtime enforcement on every call
  • Ed25519 signed, tamper-evident ledger
  • Self-host SDK and audit logs in your environment
  • Policy-as-code, JSON-configured (OPA bundles v1.1) · git-versioned
  • MIT open source · vendor-neutral by structure
  • Auditor-ready evidence for SOC 2 · HIPAA · EU AI Act
Locked-in alternatives
  • Tied to one model provider (Promptfoo → OpenAI)
  • Pre-deployment only · snapshots in CI
  • Self-reported logs · unsigned
  • Cloud-only (Azure-native, Foundry-locked)
  • Console-based rules · no version control
  • Proprietary · governed by a frontier lab
  • Screenshots + CSVs at audit time
Pricing plan

Start free. Stay free.

The SDK is MIT, forever. Pay us when you need hosted distribution, BAA, or signed forensic export.

OSSMIT · forever
$0
Self-host · unlimited agents

Everything an independent dev needs to govern production agents.

  • Full SDK · Python today · TypeScript v0.2 · Go on roadmap
  • Policy engine · Policy-as-code, JSON-configured (OPA bundles v1.1)
  • Local signed ledger
  • Community Slack
TeamMost popular
$1,500/mo
Per 5 agents · $250 each after

For teams running agents in production across 2+ providers.

  • Everything in OSS
  • Hosted control plane · dashboard
  • Slack · PagerDuty approvals
  • SOC 2 + HIPAA evidence export
  • 7-day signed ledger retention
Start free trial
EnterpriseAnnual · BAA
$75k/yr
From · volume + SLA tiers

Self-hosted, SSO, BAA, dedicated compliance engineer.

  • Everything in Team
  • Self-hosted in your VPC
  • Dedicated compliance engineer
  • Signed BAA · DPA · custom SCCs
  • Unlimited retention · WORM storage
✓ Guarantee

If a Praxel-governed agent action is ever flagged as unauditable in your SOC 2 or HIPAA audit, we refund 12 months of fees and pay your auditor up to $25,000 to re-review.

Common questions

Skeptical? Good. So are we.

The ones we get most from platform leads, CISOs, and the senior engineer in the room with arms crossed.

What's the latency cost of wrapping every call?

Sub-50ms policy evaluation in-process. No sidecar, no proxy hop. Full benchmark suite shipping with v1.0.

How is this different from Promptfoo?

Promptfoo tests agents in CI before they ship. Praxel governs them after. Testing is a snapshot. Production is a stream. Also — OpenAI owns Promptfoo now. The neutrality question matters in a regulated environment.

Does this work with my existing orchestrator?

Yes. We wrap at the model-client layer, below LangChain, LlamaIndex, CrewAI, or your own orchestration. If it eventually calls an HTTP model endpoint, Praxel sees it.

Where do audit logs live?

Wherever you want. Local filesystem, your S3 bucket, Snowflake, BigQuery, or a WORM archive. We never see your events unless you explicitly enable the hosted plane.

What happens if Praxel goes down?

The SDK is in-process. There is no SaaS dependency in the critical path unless you opt into hosted approvals. Policies are cached locally and evaluated offline.

Ready to ship?

Govern the stream.
Not the snapshot.

Install the SDK in three lines. Sign every call. Hand your auditor signed event exports instead of screenshots.