Tacit Testbench is a technical assessment platform built around modern engineering workflows. Candidates solve practical coding tasks in their own IDE, using their own AI agents and tools. Teams can review the submitted code, test results, development history, and recorded explanations together.
Built for recruiting agencies, startups, and technical teams, Tacit supports the full assessment lifecycle: creating challenges, inviting candidates, running interviews, evaluating submissions, and managing evidence across separate client companies.
Teams can build an assessment around a realistic engineering task, complete with a starter codebase, instructions, scoring criteria, and public and hidden tests. Before publishing, they can validate those tests against reference and partial solutions. Candidates receive an invitation, connect the CLI with a pairing code, and download their assessment workspace. They can run public tests repeatedly through the CLI; execution happens in an isolated environment, so local dependency or runtime problems do not prevent testing.
During the assessment, Tacit captures code changes, checkpoints, Git activity, and supported AI-agent conversations.
After submitting, candidates can inspect their submitted code, use a sandbox terminal to prepare, and record a technical walkthrough explaining their implementation and decisions.
Then, reviewers get a consolidated view of the submission, test outcomes, code history, recordings, transcripts, and AI-assisted feedback, giving them evidence to discuss and assess the candidate's work.
Features
- Flexible assessment authoring: Create assessments through the website, CLI, or MCP, or import a structured assessment package.
- Complete test design: Include starter code, public and hidden tests, scoring criteria, and reference and partial solutions for validation.
- Local IDE workflow: Candidates authenticate with a pairing code and work locally through a CLI supporting macOS, Windows, and Linux.
- Isolated test execution: Run repeatable public tests and final grading in managed sandboxes, independent of the candidate's local environment.
- Code history and replay: Review file changes, diffs, checkpoints, and Git activity throughout the assessment.
- AI-agent visibility: Capture supported coding-agent conversations alongside the code they helped produce.
- Recorded technical walkthroughs: Candidates explain their decisions through timed spoken follow-ups, with access to their submitted code and a preparation terminal.
- Evidence-based evaluation: Combine deterministic test results with transcripts and AI-assisted technical feedback for human review.
- Agency and company workspaces: Manage client companies, roles, assessment stages, staff permissions, candidate accounts, and invitations.
- Recovery and retention controls: Support reconnection, resumable uploads, deadline submission, bounded workspace cleanup, and applicant archive, restart, and deletion workflows.