v2.4 ships trace replay. Rerun any production run, step by step.

v2.4 ships trace replay. Rerun any production run, step by step.

SEE EVERY STEP YOUR AGENTS TAKE.

[00] AGENT RUNTIME

Trace, replay and version every run in production. One view for what ran, why it ran and what it cost.

No credit card

·

10k free runs/mo

·

SOC 2 Type II

app.meridel.dev/runs/tr_8f2a91

RUN tr_8f2a91● LIVE
TIMETYPESTEPSTATUSDUR

IN PRODUCTION AT

Halden Freight

Quarry Health

Tessel

Corvid Analytics

Halden Freight

Quarry Health

Tessel

Corvid Analytics

Halden Freight

Quarry Health

Tessel

Corvid Analytics

Halden Freight

Quarry Health

Tessel

Corvid Analytics

AGENTS THAT ACTUALLY SHIP.

[00] AGENT RUNTIME

The runtime for autonomous agents that plan, reason and take action in production — with every step traced, replayable and versioned.

No credit card

·

10k free runs/mo

·

SOC 2 Type II

P50 LATENCY

142ms
142ms

UPTIME

99.95%
99.95%

RUNS/SEC

3,812
3,812

IN PRODUCTION AT

Halden Freight

Quarry Health

Tessel

Corvid Analytics

Halden Freight

Quarry Health

Tessel

Corvid Analytics

Halden Freight

Quarry Health

Tessel

Corvid Analytics

Halden Freight

Quarry Health

Tessel

Corvid Analytics

[01] HOW IT WORKS

FROM PROMPT TO PRODUCTION.

Three stages, one runtime. Nothing to stitch together.

STEP 01

DEFINE.

Describe the job, the tools it can call and the rules it can’t break. Typed, versioned and reviewed like code.

STEP 02

OBSERVE.

Every step is traced: inputs, outputs, tokens, latency and cost. Replay any run exactly as it happened.

STEP 03

SCALE.

Promote a version when evals pass, roll back in one click when they don’t. Capacity follows traffic.

[02] PRODUCT

AGENTS AS CODE. TRACED BY DEFAULT.

Define agents in plain code. Every run is traced, costed and versioned with no extra setup.

TRACE / RUN_8F2C41 / SUPPORT-TRIAGE

LIVE

7 STEPS · 3.86S · $0.012

REPLAY AVAILABLE

TRACE / RUN_8F2C41 / SUPPORT-TRIAGE

LIVE

7 STEPS · 3.86S · $0.012

A — EVERY STEP

Model calls, tool calls, retries and policy checks, in the order they ran.

B — EVERY COST

Tokens, latency and spend per step, rolled up per run and per version.

C — EVERY VERSION

The exact prompt, tools and model each step ran against.

[03] FEATURES

BUILT FOR THE SECOND MONTH.

The demo is the easy part. These are the parts that matter once real traffic shows up.

01

TYPED TOOLS

Tools are declared with schemas, so bad arguments fail at the edge instead of three steps later.

02

FULL TRACES

Inputs, outputs, tokens, latency and cost for every step, searchable for 90 days.

03

DETERMINISTIC REPLAY

Re-run any production trace with its original inputs pinned, then fork it to test a fix.

04

POLICY GUARDRAILS

Step limits, spend caps and allow-lists that hold no matter what the model decides.

05

VERSIONED DEPLOYS

Every change ships as a version. Promote on passing evals, roll back in one click.

06

ANY MODEL, ANY CLOUD

Swap providers behind one interface, or run the runtime inside your own VPC.

[04] IN PRODUCTION

NUMBERS WE PUBLISH.

Measured across every workspace over the last 30 days.

STEPS TRACED / MONTH

1.2B

STEPS TRACED / MONTH

1.2B

STEPS TRACED / MONTH

1.2B

P50 STEP OVERHEAD

142ms
142ms

UPTIME, TRAILING 12 MO

99.95%
99.95%

RUNS / SEC AT PEAK

3,812
3,812

RUNS / WEEK · LAST 12 MONTHS

THIS QUARTER

Your numbers, on your dashboard, from the first run.

[06] INTEGRATIONS

PLUGS INTO WHAT YOU RUN.

Models, data, observability and workflow tools. Keep the stack you already have.

ALL

ALL

Models

Models

Data

Data

Observability

Observability

Workflow

Workflow

Missing one? The SDK wraps any HTTP API in about twenty lines.

Missing one? The SDK wraps any HTTP API in about twenty lines.

[08] PRICING

PAY FOR RUNS, NOT SEATS.

Every plan includes full tracing. Annual billing saves 20%.

MONTHLY

MONTHLY

ANNUAL −20%

ANNUAL −20%

STARTER

$0

/MO

For prototypes and side projects. 10,000 runs a month, no card needed.

+

Step-level tracing

Every model call, tool call and retry, with inputs and outputs.

PRO

MOST TEAMS

$39

/MO

For teams running agents in production. 250,000 runs included, then $0.40 per thousand.

+

Step-level tracing

Every model call, tool call and retry, with inputs and outputs.

+

Eval suites in CI

+

90-day trace retention

Enterprise retention is configurable up to 7 years.

ENTERPRISE

CUSTOM

For regulated workloads. VPC deployment, custom retention and a 99.9% SLA.

+

Step-level tracing

Every model call, tool call and retry, with inputs and outputs.

+

Eval suites in CI

+

90-day trace retention

Enterprise retention is configurable up to 7 years.

+

SSO, SCIM and audit log

SOC 2 TYPE II · Audited annually

GDPR · EU data residency

99.9% uptime SLA · Enterprise

[09] FAQ

ASKED BEFORE YOU ASK.

Security first, because that’s where the hard questions start.

Do you train models on our data?

Are you SOC 2 Type II certified?

Can we run it in our own cloud?

What counts as a run?

[10] GET STARTED

FIRST RUN IN FIVE MINUTES.

NOW SHIP YOURS.

No credit card

·

Cancel anytime

·

A human replies within one business day

[10] GET STARTED

FIRST RUN IN FIVE MINUTES.

NOW SHIP YOURS.

No credit card

Cancel anytime

A human replies within one business day

[10] GET STARTED

FIRST RUN IN FIVE MINUTES.

NOW SHIP YOURS.

No credit card

·

Cancel anytime

·

A human replies within one business day

[ FRAMER TEMPLATE ]

Get Meridel.

Designed by Zavati Studio. Remix it for your own AI startup in an afternoon.

[ CUSTOM SETUP ]

$249

Template included. I swap in your colours, fonts and copy. Live in 3 days.

Bigger changes or more languages? Ask for a quote.

[ FRAMER TEMPLATE ]

Custom setup from $249, template included. Colours, fonts and copy, live in 3 days. More? Ask.

Create a free website with Framer, the website builder loved by startups, designers and agencies.