8
;
8
}
%
2
|
9
@
?
2
6
\
^
:
\
~
^
%
0
=
&
*
&
8
0

3
9
-
/
~
2
[
/
[
0
9
3
9
0
8
}
+
1
[
#
*
^
6
-
-
}

5
~
-
&
&
:
~
1
<
6
0
{
8
>
-
6
?
&
\
7
>
#
]
?
3
*

;
5
;
2
1
*
$
6
0
|
<
3
8
*
=
8
-
}
|
\
{
2
&
2
[
{

0
;
&
-
8
?
6
{
9
6
0
0
>
>
1
=
7
3
*
/
-
^
|
0
&
8

%
3
{
6
+
$
5
@
7
*
-
]
^
*
?
;
<
]
^
+
2
0
\
=
|
+

+
&
\
1
%
}
>
5
?
3
|
\
>
2
:
;
-
>
1
-
=
$
8
~
{
8

-
{
-
~
5
@
&
8
+
7
&
}
4
/
0
5
?
<
}
{
?
8
>
7
~
&

7
}
[
$
~
2
5
^
]
7
%
3
:
5
;
&
}
*
0
@
/
*
1
9
~
7

+
;
~
@
[
7
0
}
7
$
1
^
}
6
|
\
5
;
|
<
}
1
~
]
&
2

^
*
:
*
4
0
~
8
8
=
=
#
{
/
0
[
5
}
0
;
<
?
=
\
^
|

|
:
1
{
<
\
6
8
1
[
+
#
;
<
1
{
:
9
<
&
<
&
&
3
>
0

3
]
^
;
[
;
]
^
0
:
#
2
*
$
#
>
/
0
5
8
&
-
}
>
~
|

?
$
}
+
#
8
#
&
8
;
\
?
:
|
6
@
}
^
0
9
]
7
$
;
8
9

~
4
:
-
|
[
\
=
/
;
^
|
\
:
8
\
4
3
=
?
%
9
1
5
&
0

2
[
-
<
:
-
1
3
+
9
$
%
/
$
8
0
+
+
{
9
[
2
\
<
9
:

{
8
0
4
4
-
-
1
<
=
|
}
3
:
8
4
6
&
>
$
4
6
{
/
4
4

0
?
0
3
2
#
1
[
;
{
6
-
+
;
1
5
[
?
#
7
*
]
5
[
7
1

@
~
|
}
<
+
9
3
-
*
?
0
#
:
7
~
?
5
{
0
4
-
0
]
$
4

|
~
{
]
7
#
&
=
0
{
=
7
>
0
7
|
*
=
;
}
<
<
=
}
+
$

}
+
$
;
7
2
7
<
$
&
7
%
?
&
3
=
]
4
:
%
@
8
%
#
2
;

[
3
$
~
|
=
@
^
%
;
0
:
*
[
4
^
1
5
$
\
%
<
|
-
[
;

\
=
{
*
{
<
3
%
*
}
^
/
&
\
9
\
3
]
&
\
/
]
5
<
5
0

<
{
0
0
0
:
^
]
#
2
%
1
7
]
>
~
4
#
6
{
>
;
*
[
\
>

}
^
7
7
0
^
9
9
1
&
1
{
>
{
5
4
6
=
2
5
2
5
~
+
2
&

:
?
~
<
?
<
#
>
0
1
4
0
#
^
[
8
9
~
?
*
9
+
&
%
;
%

@
+
-
7
:
5
<
&
7
=
\
<
%
=
{
[
0
@
=
~
%
]
;
<
&
/

:
5
8
2
@
/
{
1
<
;
5
&
<
2
-
8
6
;
[
8
8
5
%
-
-
]

2
&
=
\
2
{
~
5
2
+
<
[
;
7
4
4
<
^
#
#
&
[
7
#
&
?

\
~
1
%
\
^
1
1
%
$
#
7
&
;
>
5
~
@
=
;
6
$
/
^
0
;

?
=
=
@
$
{
=
3
=
*
8
6
3
/
1
#
=
1
:
&
}
:
>
@
?
4

@
*
@
@
\
0
>
&
/
2
-
2
^
~
&
|
^
2
*
8
0
3
&
*
[
2

2
&
:
2
[
2
@
]
}
~
?
=
$
6
0
~
/
3
4
0
3
1
^
<
2
<

|
&
}
6
}
|
<
3
0
&
1
{
8
@
%
3
]
:
~
{
|
@
0
8
|
{

=
]
@
|
/
8
~
0
\
[
]
&
7
5
6
:
0
|
8
&
0
?
~
-
2
]

9
1
8
&
6
^
;
9
]
+
}
<
;
5
@
4
{
|
^
%
8
8
+
|
]
;

5
9
+
*
$
0
3
<
;
:
9
/
+
0
7
-
|
$
0
4
*
1
:
|
;
@

*
5
~
&
9
^
[
&
=
3
6
=
$
+
<
}
%
@
:
6
>
$
@
{
4
#

[
5
=
+
*
<
]
~
@
^
>
2
2
>
:
]
[
-
}
&
-
+
@
4
&
;

6
@
\
\
[
;
7
^
8
&
[
3
~
-
1
^
|
=
\
=
@
8
:
3
[
&

5
{
@
<
^
%
;
2
-
2
{
0
0
;
1
+
%
-
\
&
2
/
8
8
?
1

8
]
0
$
6
^
8
>
0
-
9
%
-
9
^
/
1
}
0
1
/
4
?
*
+
8

#
|
%
[
0
*
%
\
+
;
}
#
^
<
>
+
\
4
2
?
{
4
:
#
&
~

1
]
=
~
0
^
&
;
[
~
2
]
@
]
&
~
/
$
\
3
^
#
^
=
3
?

%
&
&
#
]
*
?
&
[
0
*
4
]
:
&
$
{
%
-
#
@
0
0
&
^
8

=
<
-
4
4
0
4
>
6
/
@
^
?
>
/
2
&
;
[
\
0
3
$
1
*
*

Run your agents
under one unified API.

Run production agents on Onefold with one OpenAI-compatible endpoint for every frontier model, per-call cost and latency tracing, and real traces you can replay as tests.

Get API key Read the docs

app.onefoldhq.com/keys

API keys

New key

Balance: $482.19
Spent this month: $133.25
Active keys: 4
Projects: 3

Name	Key	Project	Limit	Spent	Last used
runway-prod	sk-ag-…4f2a	runway	$200 / mo	$84.21	2m ago
tasks-prod	sk-ag-…9c11	tasks	$100 / mo	$41.09	14m ago
atlas-dev	sk-ag-…20e7	atlas	None	$6.83	3h ago
local	sk-ag-…7b55	runway	$20 / mo	$1.12	2d ago

Name

Key

Project

Limit

Spent

Last used

runway-prod

sk-ag-…4f2a

runway

$200 / mo

$84.21

2m ago

tasks-prod

sk-ag-…9c11

tasks

$100 / mo

$41.09

14m ago

atlas-dev

sk-ag-…20e7

atlas

None

$6.83

3h ago

local

sk-ag-…7b55

runway

$20 / mo

$1.12

2d ago

runway / traces / tr_8f2a4c

runway.forecast_query

sess_7c2 · trace 3 of 4

Duration: 3.84s
Cost: $0.0536
Tokens: 14,208
Spans: 6

plan.intent340ms$0.0021
tool.fetch_txns890mstimeout
tool.fetch_txns210msretry · ok
retrieve.policy180ms8 docs
reason.projection1.92s$0.0448
compose.summary480ms$0.0067

reason.projection llm

model: claude-sonnet-4-5
ttft: 312ms
tokens: 8,140 in / 602 out

Seeded from tr_8f2a4c

Run all

claude-opus-5

How long is our runway?

Net burn is $18.4k a month. At the current balance you have 14 months of runway, ending Nov 2027.

1.84s · 312ms ttft · 8,140 → 602 · $0.0448

claude-sonnet-5

How long is our runway?

Runway is roughly 14 months on $18.4k net burn.

0.91s · 198ms ttft · 8,140 → 214 · $0.0076

claude-haiku-4-5

How long is our runway?

14 months.

0.38s · 96ms ttft · 8,140 → 41 · $0.0011

vs claude-opus-5 · cost −83% · latency−51% · output −64%

Run

One baseURL, every frontier model
Automatic failover across providers
Per-key spend limits and budgets
Streaming passthrough, no added latency

Trace

Full span trees, tool calls included
Cost, tokens, and latency per step
Sessions across multi-turn runs
OpenTelemetry native, any framework

Experiment

Compare models side by side
Seed any run from a real trace
Real cost, not estimates
Every run traced automatically

Write-ups

ContextMeasured through Onefold

A 72:1 input-to-output ratio in a five-step agent run

Sixty thousand tokens in, eight hundred out. Where the rest of it went, step by step.

Coming soon

AgentsMeasured through Onefold

What 3 coding agents cost to finish the same task

Cursor, Claude Code and Cline, one ticket, one repo.

Coming soon

Tool callsMeasured through Onefold

The cheapest model that still calls the right tool

Coming soon

RoutingMeasured through Onefold

One model, 4 providers: does the endpoint change the answer?

Same weights, different serving stack. Latency and throughput measured per endpoint.

Coming soon

FrontendMeasured through Onefold

Is Kimi better at building UI than the top 3 models?

Coming soon

CostMeasured through Onefold

Accuracy per dollar, not accuracy

The ranking nobody publishes, because it needs first-party pricing.

Coming soon

Run your agentsunder one unified API.

Run

Trace

Experiment

Write-ups

A 72:1 input-to-output ratio in a five-step agent run

What 3 coding agents cost to finish the same task

The cheapest model that still calls the right tool

One model, 4 providers: does the endpoint change the answer?

Is Kimi better at building UI than the top 3 models?

Accuracy per dollar, not accuracy

Run your agents
under one unified API.