Clay AI agent benchmark: OpenCode with Clay

With Clay, OpenCode with Kimi K3 found 93 of 100 verified work emails in every run at a median of 5.1 minutes, the fastest of seven AI coding agents on that task, for as little as 337.5 Clay credits, about 3.4 per contact. On job changes it scored 93 to 95 in every run at about 14 credits per person, catching 42 of the 45 people who had moved each time.

By The Prompting Company · Snapshot November 5, 2026

What can OpenCode do with Clay?

We ran OpenCode with Kimi K3 (OpenCode Kimi below) through five GTM data tasks with Clay, three runs each. Clay credits are measured as the change in the Clay workspace credit balance.

OpenCode Kimi was quick and light on model cost. When it ran one Clay routine per record, it finished 100 records in minutes with scores among the best in the benchmark.

What OpenCode does well with Clay

Work emails, fast. OpenCode Kimi ran Clay's Work Email routine for each contact and found the same 93 of 100 in every run, in 5.1, 3.7 and 8.7 minutes. Its first run sent one test contact, then two batches of 50, and spent 337.5 credits. Model cost was 22 to 37 cents per run.

Job changes. OpenCode Kimi ran Clay's Enrich Person and Find Contact Details once per LinkedIn profile and scored 93, 93 and 95 for 1,404 to 1,406 credits, about 14 per person. Its score moved only 2 points across the three runs, and each run took 8.6 to 13.5 minutes.

Company data with a check in Clay. In its first run, OpenCode Kimi checked about 40 ambiguous company names against Clay's company search before enriching. It scored 98 on domains, 99 on HQ, 99 on headcount and 70 on funding for 324 credits, about 3.2 per company.

Finding the CMO. In runs 2 and 3, OpenCode Kimi found 72 and 73 of 100 CMOs with Clay's people search and Work Email, clearing the pass line both times.

A target list for 7 credits. Its third run used Clay's Find Companies search with the "Public Company" company type, a US headquarters and software and IT industries, and built a 100-company list for 7 credits in 3.7 minutes.

OpenCode results with Clay, by task

Primary score out of 100 for each of three runs, with the average; Clay credits and minutes are per run. Company data is the average of domain, HQ, headcount and funding.

Task	Runs	Avg	Clay credits	Minutes
Work emails	93 · 93 · 93	93.0	337.5 · 343.5 · 373.5	5.1 · 3.7 · 8.7
Company data	91.5 · 82.5 · 0*	58.0	324 · 337 · 1,531	31.0 · 9.5 · 60.6
CMO	0* · 72 · 73	48.3	1,665.5 · 450 · 631	60.6 · 17.2 · 27.6
Job changes	93 · 93 · 95	93.7	1,404 · 1,404 · 1,406	8.6 · 13.5 · 10.3
Target list	44 · 66 · 66	58.7	101 · 359 · 7	9.9 · 6.4 · 3.7
All five		70.3	3,560 (sum of averages)	276.4 total

*This run reached the 60-minute limit and counts as 0.

Minutes are measured from run start to grading. Reported model cost was $13.88 over 13 runs, about $1.07 per run.

The best setup for OpenCode with Clay

Install the Clay Agent Plugin. It ships Clay's skills to OpenCode up front, including the habits that keep credit use down. Setup is one prompt, about 15 minutes, and there is no additional cost to use it.
Let Clay's API guide the inputs. Clay's API tells an agent exactly what each routine needs, so OpenCode Kimi self-corrected in one call.
Send one test item, then large batches. OpenCode Kimi's two batches of 50 cost exactly the same as ten batches of 10, and fewer batches meant fewer polling rounds.
Name the Clay method in the prompt, so the agent goes straight to the low-cost path:

Task	Clay method	Result in the benchmark
Work emails	Work Email once per contact	OpenCode Kimi 93 in 3.7 to 8.7 minutes
Job changes	Enrich Person and Find Contact Details once per profile	OpenCode Kimi 93 to 95
Company data	Check candidates in company search, then enrich	OpenCode Kimi 98 on domains
CMO	Find People to choose, then one Work Email	Up to 82 for under 3 credits per company
Target list	Find Companies, confirm listing, US HQ and software	82 to 94 qualifying companies

Factory runs the same Kimi K3 model; for how the two harnesses compare with Clay, see "Does the agent harness matter when using Clay with the same model?".

How we tested

Seven agents (Claude Code with Opus 5.5 and Sonnet 5, Codex with GPT-6 Sol, Pi with GPT-5.6 Luna, Grok CLI with Grok 4.6, OpenCode with Kimi K3, Factory with Kimi K3) ran each of five tasks three times, one agent at a time, in a fresh sandbox with the same Clay key, the same 19 saved Clay routines and a 60-minute limit. Runs interrupted by the test platform were rerun. Pass lines: 70 for emails; 90 for domain and HQ and 70 for headcount and funding; 70 for CMO name and title and 60 for email; 80 for job changes; 70 qualifying companies plus a full list of 100 for the target list. Tasks and graders: github.com/promptingcompany/clay-bench.

Try it yourself

Install the Clay Agent Plugin for OpenCode, or give it a Clay API key and the Clay Public API docs. Add the method for your task from the table above as one line in the prompt. Agent work uses the same credits as in-product work, and an enrichment that returns no result is not charged (Clay pricing).

Common questions

What can OpenCode do with Clay? OpenCode with Kimi K3 found 93 of 100 work emails in every run at a median of 5.1 minutes, the fastest of seven agents, and scored 93 to 95 on job changes in every run.

How many Clay credits does OpenCode use per record? About 3.4 per contact for work emails and about 14 per person for a job change check.

What does OpenCode cost to run with Clay? About $1.07 of model cost per reported run, and 22 to 37 cents on work emails, plus Clay credits.

Does it cost extra to use Clay from OpenCode? No. There is no additional cost to use the Clay Agent Plugin; agent work uses the same credits and actions as in-product work.