System OneENGINE

CAPABILITY DIRECTORY

Find the decision
your agent repeats.

Each listing says what the agent hands over, how the research rates that handoff, and what the recipe sample costs as input. Open the article for the example, the boundary, and a calculator.

Search guides and findings too ↗Change the cost assumptions ↗

The dollar figure uses 10,000 uses of the recipe sample. Jev is the published $0.042 per million input list rate. The other figure uses a $3 per million example rate. Token count is the sample length divided by four. It is a planning estimate, not a tokenizer measurement. Output charges, retries, hosting, and task quality are not included.

Agent flow

8 capabilities

Decision

Ask or proceed

Identify missing information that blocks a concrete next step.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.06Jev list rate. About $3.99 at the $3 per million example rate. About 133 input tokens in the sample.

Decision

Choose the next tool

Choose among a small set of tools with explicit capabilities.

Research value
MediumEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.39 at the $3 per million example rate. About 113 input tokens in the sample.

Decision

Choose who should handle a request

Send a request to deterministic code, a specialist model, or a person.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.10Jev list rate. About $6.99 at the $3 per million example rate. About 233 input tokens in the sample.

Decision

Interrupt or stay quiet

Decide whether a background result deserves attention now.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.03Jev list rate. About $2.16 at the $3 per million example rate. About 72 input tokens in the sample.

Decision

Match a command to a request

Map an informal request to an available command ID.

Research value
MediumEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.76 at the $3 per million example rate. About 92 input tokens in the sample.

Decision

Recognize the current intent

Separate a question, a change request and a status request.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.82 at the $3 per million example rate. About 94 input tokens in the sample.

Decision

Review a proposed tool call

Classify a proposed tool call as clear or caution before anything runs.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.10Jev list rate. About $6.99 at the $3 per million example rate. About 233 input tokens in the sample.

Decision

Understand a tool result

Recognize success, failure or missing evidence without another long response.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.30 at the $3 per million example rate. About 110 input tokens in the sample.

Classification

2 capabilities

Catalog selection

Navigate a decision tree

Use jev-tree to choose a known category with a bounded call budget.

Research value
MediumEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.03Jev list rate. About $2.49 at the $3 per million example rate. About 83 input tokens in the sample.

Decision

Route a support request

Choose a queue and flag an explicit time-sensitive problem.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.08Jev list rate. About $5.85 at the $3 per million example rate. About 195 input tokens in the sample.

Coding

2 capabilities

Decision

Compare issue reports

Flag whether two reports describe the same observed failure.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.97 at the $3 per million example rate. About 99 input tokens in the sample.

Decision

Route a code change

Choose the right review path from a short change description.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.88 at the $3 per million example rate. About 96 input tokens in the sample.

Computer use

6 capabilities

Decision

Check a browser outcome

Ask whether the final screen supports the result the user requested.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.94 at the $3 per million example rate. About 98 input tokens in the sample.

Decision

Check a form against the task

Compare requested values with a visible form before the agent submits it.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.76 at the $3 per million example rate. About 92 input tokens in the sample.

Decision

Choose the next browser operation

Choose a small next step from the current page state and task.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.07Jev list rate. About $5.10 at the $3 per million example rate. About 170 input tokens in the sample.

Decision

Match a visible control

Choose a control ID from descriptions captured by your browser tool.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $3.06 at the $3 per million example rate. About 102 input tokens in the sample.

Decision

Read what changed on screen

Classify observed progress after an action without inventing a result.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.06Jev list rate. About $4.29 at the $3 per million example rate. About 143 input tokens in the sample.

Decision

Return a stuck browser task

Choose whether to wait, observe again or return control to the main agent.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.78 at the $3 per million example rate. About 126 input tokens in the sample.

Context

5 capabilities

Decision

Find a useful preference

Distinguish a reusable explicit preference from incidental conversation.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $3.12 at the $3 per million example rate. About 104 input tokens in the sample.

Decision

Flag how a passage relates to a task

Separate useful evidence, contradiction and embedded instructions.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.06Jev list rate. About $4.02 at the $3 per million example rate. About 134 input tokens in the sample.

Decision

Keep useful context

Evaluate candidate context before adding it to a longer prompt.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $3.00 at the $3 per million example rate. About 100 input tokens in the sample.

Decision

Select a useful source snippet

Select a source ID using the meaning of the request.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $3.03 at the $3 per million example rate. About 101 input tokens in the sample.

Decision

Spot stale evidence

Identify whether a supplied fact needs a fresh lookup.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.03Jev list rate. About $1.95 at the $3 per million example rate. About 65 input tokens in the sample.

5 capabilities

Decision

Choose the next corpus search step

Choose among retrieval expansion, keyword filtering, excerpt reading, reformulation or conclusion.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.09Jev list rate. About $6.69 at the $3 per million example rate. About 223 input tokens in the sample.

Decision

Reformulate agent search query

Choose how to refine a search query when BM25 or grep yields zero or noisy matches.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.07Jev list rate. About $5.10 at the $3 per million example rate. About 170 input tokens in the sample.

Decision

Select top candidate document

Pick the most promising candidate document from BM25 retrieval to stage or inspect first.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.06Jev list rate. About $4.14 at the $3 per million example rate. About 138 input tokens in the sample.

Decision

Triage grep match lines

Separate high-signal matches containing relevant evidence from routine boilerplate.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.60 at the $3 per million example rate. About 120 input tokens in the sample.

Decision

Verify claim against candidate passage

Compare a factual claim with a concrete line range or passage retrieved from the corpus.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.85 at the $3 per million example rate. About 95 input tokens in the sample.

Design

1 capability

Decision

Match a color palette

Choose an approved palette that fits a written brief.

Research value
MediumEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.36 at the $3 per million example rate. About 112 input tokens in the sample.

Experiences

1 capability

Dialogue

Choose a natural moment

Let an optional adviser choose an eligible recorded reaction or silence.

Research value
MediumEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.03Jev list rate. About $1.95 at the $3 per million example rate. About 65 input tokens in the sample.

Operations

2 capabilities

Decision

Choose a diagnostic test

Rank proposed checks using current failure evidence.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.45 at the $3 per million example rate. About 115 input tokens in the sample.

Log triage

Prioritize log inspection

Use jevlogs to separate routine events from records that deserve investigation.

Research value
LowEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.02Jev list rate. About $1.29 at the $3 per million example rate. About 43 input tokens in the sample.

Research

1 capability

Decision

Screen a document against criteria

Check an abstract against explicit inclusion criteria.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.57 at the $3 per million example rate. About 119 input tokens in the sample.

Selection

3 capabilities

Decision

Compare two item records

Flag whether two descriptions likely refer to the same item.

Research value
MediumEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $3.06 at the $3 per million example rate. About 102 input tokens in the sample.

Decision

Match an item to a description

Choose a supplied asset, template or catalog item by its description.

Research value
HighEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.33 at the $3 per million example rate. About 111 input tokens in the sample.

Decision

Select values for known fields

Map a request to allowed field values without generating arguments.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.24 at the $3 per million example rate. About 108 input tokens in the sample.

Verification

7 capabilities

Decision

Check a citation in context

Compare a claim with the supplied source passage.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.48 at the $3 per million example rate. About 116 input tokens in the sample.

Decision

Check a claim

Compare a short claim with the evidence already collected.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.58 at the $3 per million example rate. About 86 input tokens in the sample.

Decision

Check a repair against its invariant

Distinguish temporary recovery from a repair that meets stated requirements.

Research value
HighEvidence confidence is moderate. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.66 at the $3 per million example rate. About 122 input tokens in the sample.

Decision

Check launcher acceptance

Batch independent checks over one observed result.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.54 at the $3 per million example rate. About 118 input tokens in the sample.

Decision

Review a draft before it is sent

Check a draft reply against the evidence, then recommend send, hold, or review.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.10Jev list rate. About $7.44 at the $3 per million example rate. About 248 input tokens in the sample.

Decision

Review a proposed refund

Check whether a refund request names an amount and looks like a duplicate charge.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.11Jev list rate. About $7.62 at the $3 per million example rate. About 254 input tokens in the sample.

Decision

Score a small rubric

Evaluate a bounded quality scale with explicit anchors.

Research value
Not ratedNo research record is pinned. Compare it with your own task.
Input estimate for 10,000 uses
$0.04Jev list rate. About $3.03 at the $3 per million example rate. About 101 input tokens in the sample.

Writing

2 capabilities

Decision

Check named writing rules

Flag vague wording and missing specifics before an agent edits a draft.

Research value
MediumEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.05Jev list rate. About $3.87 at the $3 per million example rate. About 129 input tokens in the sample.

Decision

Label a text block

Assign a known structural label to supplied text.

Research value
MediumEvidence confidence is low. This is a research judgment, not a measured return.
Input estimate for 10,000 uses
$0.04Jev list rate. About $2.85 at the $3 per million example rate. About 95 input tokens in the sample.