New evaluation
Give an agent a task. Traceva verifies the outcome and shows the evidence.
Traceva beta
Checking your session…
Recent evaluations
Configuration
What should the agent accomplish?
Waiting to connect.
This MVP verifies the seeded Apollo and Alice outcome, not arbitrary tasks or websites.
Discovered tools
0
of 4 ready
WebMCP contract
Detecting schema…
Traceva reads the live add_member definition.
Connect to all four tools to begin.
Cross-model comparison
Same task. Same tools. Different agent.
Recommended repair
Make both identifiers explicit
The ambiguous add_member contract exposes one generic id even though the operation requires a project and a member.
− id
+ projectId + memberId
Failure reproduced
Reloads the target and rediscovers its real WebMCP tools.
Before
FAIL 0/100 Ambiguousid contract
After
PASS 100/100 ExplicitprojectId + memberId
The same model and task now reach the verified state.
Verified result
—
Score
0
/100
Evidence
Agent trajectory
Connection details Tool schemas, state controls, and native execution check
Browser integration check
Discover tools before running the execution check.
No check run yet.