A Gartner-style read on where Tenki sits in the agent-era compute field — across CI runners, AI code review, and agent sandboxes. Tenki is the only player spanning all three on one Firecracker substrate, on compute Luxor owns. Every DevRel initiative in this plan attaches to a strength to amplify, a weakness to close, an opportunity to capture, or a threat to neutralize — the codes below (S/W/O/T) are referenced throughout the strategy.
S1 Drop-in GitHub Actions replacement (one runs-on change + Migration Wizard) — lowest switching friction in the categoryS2 Firecracker microVM per-job isolation on owned bare metalS3 Three products on one substrate (runners + review + sandbox)S4 Captive low-cost compute via Luxor (mining + GPU build-out)S5 AI Code Reviewer leading recall (68.9%, ~2× next best)S6 Sandbox = agent-era infra (Claude Code / Codex), <100ms bootS7 Profitable parent, SOC 1/2 Type IIS8 ~Weekly shipping cadence (81 releases)W1 Few public customer references (warm / crypto network)W2 AI reviewer precision low (29.9%) — trust problemW3 No developer community / brand; empty blogW4 Crypto / Bitcoin-mining association may deter some devsW5 Not open-source — weaker contribution flywheelW6 Pricing in flux; GitHub price cut compresses savingsW7 Windows runners marketed but not GAW8 Brand unknownO1 GitHub Actions pricing backlash → migration momentO2 AI-coding-agent explosion → Sandbox demandO3 No dominant AI-code-review brand yetO4 Agent-era CI convergence (runner + review + sandbox)O5 "Compute as a commodity" + Luxor supply narrativeO6 Co-marketing with Anthropic / OpenAI / CursorO7 OSS-first program → grassroots adoptionT1 Better-funded pure-plays (Blacksmith $13.5M GV, Depot $10M)T2 GitHub cuts prices / ships own AI review (Copilot)T3 AI reviewers (CodeRabbit, Greptile, Cursor Bugbot) win mindshareT4 Sandbox rivals (E2B, Modal, Daytona) own agent-infra mindshareT5 Crypto-winter / Luxor association riskT6 CI runner commoditizationT7 Precision / trust gap erodes AI-review credibilityZooming in on one product — Tenki's agent Sandbox — against the purpose-built sandbox field (E2B, Daytona, Modal, Blaxel, Runloop, Contree) and the open-source primitives commoditizing the layer beneath them. Codes use an SB- prefix so they don't collide with the page-level S/W/O/T above. Sourced from outside-in market research; competitor claims are public.
SB-S1 Firecracker microVM = the documented security sweet spot for running untrusted, prompt-injectable LLM code (dedicated kernel per sandbox, container-like boot)SB-S2 Native ADE desktop app — rare in a field that is overwhelmingly SDK/API-onlySB-S3 Sandbox sits on the same substrate as the runners and the AI reviewer — one integrated write-run-review loop, not three vendors stitched togetherSB-S4 Owned-compute pricing via Luxor — a structural cost floor no VC-funded pure-play can matchSB-S5 First-class Claude Code + Codex support — meets agent builders on the harnesses they already useSB-S6 Persistent volumes + snapshot/restore — fits long-running and resumable agents, not just one-shot executionSB-W1 Behind on mindshare and ecosystem vs. E2B (200M+ sandboxes run) and Daytona's brandSB-W2 No public GPU offering — Modal (A100/H100), Daytona, and Northflank (H100s) all expose GPU; Tenki does not yetSB-W3 Onboarding / first-run friction — a real setup snag was hit while testing the ADESB-W4 Smaller community and contributor base than the incumbentsSB-W5 No published sandbox benchmark yet — nothing concrete to anchor a boot-time or isolation claimSB-O1 Most "sandboxes" are an afterthought of a broader cloud business — room to own genuinely sandbox-first mindshareSB-O2 Long-running / self-improving / persistent agents are underserved — rivals optimize for short-lived execution; claim that categorySB-O3 MCP-native integration is where agent tooling is standardizing — lead with it rather than bolting it on laterSB-O4 Easy local/native setup is a gap (Daytona is hard to self-host) — the native ADE can own the low-friction on-rampSB-O5 Harness-agnostic positioning — Claude Code, Codex, and Cursor — instead of betting on a single agent frameworkSB-T1 Layer D — Claude Managed Agents (Apr 2026) collapses model + sandbox + harness + state into one API, eliminating the Layer-B buying decision for many teamsSB-T2 Well-funded / specialized pure-plays: E2B, Modal (50K+ concurrency), Blaxel (YC X25, $7.3M), Daytona, Runloop (SWE-bench focus)SB-T3 Contree — Nebius's git-native branching sandbox — a differentiated architecture rivals can't quickly retrofitSB-T4 OSS commoditization of the layer: Alibaba OpenSandbox, Tencent Cube, ByteDance AIO, MicrosandboxSandbox takeaway. Tenki should specialize on the long-running-agent wedge (SB-O2), the integrated write-run-review loop (SB-S3), and a sandbox-first brand (SB-O1) — because point-tool parity won't beat E2B/Daytona on their own turf, and Layer D (SB-T1) is coming for the generic, short-lived case.
Axes: developer mindshare (low → high) against breadth of the agent-era loop covered (a single point tool → the full write-run-review loop). The point tools cluster low-breadth; the incumbent with mindshare (GitHub Actions) sits mid-breadth. Tenki is alone in the high-breadth band — its job is to earn the mindshare to match.
High breadth: Tenki is the only vendor spanning CI runners + AI review + agent sandbox on one substrate (the convergence story pure-plays can't match). Building mindshare today: no community, empty blog, almost no public logos. The 12-month DevRel program is engineered to move Tenki right — into the high-mindshare / full-loop band — before a better-funded pure-play defines the category for it.
CI runners, AI code review, and agent sandboxes are three markets today. In the agent era they collapse into one loop — and Tenki is the only name that appears in all three columns of the field.
| Vendor | Category | What it is | Funding / note |
|---|---|---|---|
| Tenki | Runners + AI review + Sandbox | Drop-in GitHub Actions runners, AI PR reviewer, and disposable agent microVMs on one Firecracker substrate | Product of Luxor (profitable, SOC 1/2 Type II) — the only vendor spanning all three |
| GitHub Actions | CI runner | The incumbent hosted-CI default Tenki replaces with one runs-on change | Microsoft; recently cut hosted-runner prices (compresses headline savings) |
| Depot | CI runner | Faster GitHub Actions runners + remote Docker/build caching | $10M Series A (Felicis) |
| Blacksmith | CI runner | Faster, cheaper GitHub Actions runners on bare metal | $13.5M (GV / YC) |
| Namespace | CI runner | High-performance runners + remote build/cache infrastructure | Well-funded runner pure-play |
| WarpBuild | CI runner | Drop-in faster/cheaper GitHub Actions runners | Runner point tool |
| BuildJet | CI runner | Performance GitHub Actions runners at lower cost | Runner point tool |
| Ubicloud | CI runner | Open-source cloud incl. cheaper GitHub Actions runners | Open-source IaaS / runner play |
| RunsOn | CI runner | Self-hosted runners in your own AWS account | Self-hosted runner point tool |
| CodeRabbit | AI review | AI PR reviewer; current adoption-mindshare leader in the category | The brand to beat on AI review |
| Greptile | AI review | Codebase-context AI PR reviewer | AI-review point tool |
| GitHub Copilot | AI review | Copilot code review native in GitHub PRs | Microsoft; platform-bundled threat (T2) |
| Cursor Bugbot | AI review | AI bug-catching review from the Cursor ecosystem | Cursor; rides editor distribution |
| Graphite | AI review | Stacked-PR workflow with AI review (Diamond) | AI-review + code-review workflow |
| Devin | AI review / agent | Autonomous coding agent that also reviews PRs | Cognition; agent-adjacent |
| E2B | Sandbox | Open-source cloud sandboxes for AI agents / code interpreters | Agent-sandbox mindshare leader |
| Modal | Sandbox | Serverless compute + sandboxes for AI / agent workloads | Well-funded compute platform |
| Daytona | Sandbox | Fast, disposable dev environments / sandboxes for agents | Agent-sandbox pure-play |
Tenki's unique position. Every name above is a point tool in exactly one column — a faster runner, an AI reviewer, or an agent sandbox. Tenki is the only vendor that spans all three — runners + AI review + sandbox — on one Firecracker microVM substrate, on compute Luxor owns. No pure-play can match the convergence story, and none of them ride a captive-compute cost moat. The strategic risk isn't the product; it's that a better-funded pure-play defines the category before Tenki earns the developer mindshare to claim it. That is the entire DevRel mandate.
Tenki has the broadest footprint in the field and a structural cost moat no VC-funded pure-play can match — but it's unknown, with almost no public references and an empty blog. Every initiative in this plan attacks a specific code above: the Benchmark Series turns the precision gap (W2/T7) into the brand; reference architectures defend the Sandbox flank (S6/T4); the Migration Engine presses the GitHub-pricing moment (S1/O1); cohorts + OSS manufacture the references (W1/W3); and ecosystem co-marketing converts incumbent threats into channels (O6/T2-T4). Win the mindshare and the loop is Tenki's to keep.