The engineering harness for AI coding agents

Your engineering harness.
Any agent. Any model.

Your developers keep Claude Code, Codex and Cursor, or use srooter's own agent and Studio app. Either way, one layer underneath gives every agent your organization's code context, standards, reviews and model policy.

$curl -fsSL srooter.ai/install | sh && srooterctl initmacOS · Windows · Linux
srooter · agent session7f3a2b · billing-svc@a1c9f2 · 3m 12s
▸add partial refunds to the billing API
HARNESS
contextmapped blast radius → payments · ledger · invoicesCortex code graph
riskHIGH · money + auth path
practicetests required before implementationyour org standard
modelglm-5.3 implements · claude-fable-5.1 reviewsby risk tier
validatebuild ✓ e2e ✓ visual ✓Validation Center
reviewcouncil · 5 seats · 1 blocker caught and fixedindependent models
✓PR #482 opened·8/8 checks·fully auditedview audit →
1 blocker caught: refund > captured amount → fixed before review

One request. Your context, standards and policy, applied automatically.

WORKS WITH
Claude CodeCodexCursoraidersrooterctlyour internal agents
RUNS ON
AnthropicOpenAIGoogleDeepSeekGLMKimilocal · Ollama
58×6 frontier models built the same app. Same result, 58× cost spread. We measure it →
THE PROBLEM

Coding agents scaled faster than the system around them.

Software engineering has become agent-first, and model providers stay heterogeneous. What is missing is the layer the organization owns, knowledge, memory, process, evaluation and model policy, that every agent runs inside.

One shared harness fixes it.

Every developer assembles their own AI setup→One harness, defined by your platform team
Each agent sees different context→The right code context per task, from a live code graph
Architecture knowledge scattered in docs→ADRs and codemaps delivered to the agent
Memory dies between sessions→Task-scoped memory that survives compaction
Models chosen ad hoc, by whoever is driving→Model chosen by task, risk and policy
Nobody can see what any agent did→Every decision validated, reviewed and recorded
Claude CodeCodexCursoraidersrooterctlyour agents
srooter harnessowned by your platform team · applied per task
KnowledgeCode graph, blast radius, ADRs, codemaps
MemoryTask-scoped state across sessions
PracticeTDD, architecture and review standards
ValidationFunctional and visual proof on every change
ReviewIndependent multi-model council
Model policyRouting, budgets, allowlists, audit
AnthropicOpenAIGoogleDeepSeekGLMKimiOllama · localyour models

Developers choose their agents. The organization owns how engineering gets done.

HOW IT WORKS

Point your agents at Srooter. Keep your tools. Gain a system.

01 · Connect

One endpoint. Nothing else changes.

OpenAI- and Anthropic-compatible. Claude Code, Codex and Cursor keep working exactly as they do today.

# one line, nothing else changes
export ANTHROPIC_BASE_URL=api.srooter.ai
02 · Apply your standards

Your context, practice and model policy, per task.

Srooter assembles the right code context, enforces your TDD, architecture and review practice, and picks the model for the task.

policy: tests-before-code
context: code-graph + ADRs
model: by task and risk
03 · Ship with evidence

Validated, reviewed, recorded.

Functional and visual validation, independent multi-model review, and a full audit trail on every change.

✓ validated · reviewed
✓ 8/8 checks
✓ audit 7f3a2b recorded
THE WALKTHROUGH

One task. The whole harness.

A developer types add partial refunds. Here is everything Srooter does before the PR opens.

The model is one step.
The harness is the other nine.

  1. 01
    Maps the blast radiusResolves the change fan-out: payments, ledger and invoices. Selects the tests that matter.Cortex code graph
  2. 02
    Assembles the architecturePulls the relevant ADRs and codemap, including the append-only ledger invariant.Codemap · ADR-041
  3. 03
    Grades the riskMoney movement on an auth path. Tier: high. An independent risk model scores it, blind to the solver.risk pipeline
  4. 04
    Requires tests firstYour TDD standard applies: failing tests before implementation.org policy · tests-before-code
  5. 05
    Picks the model for the tierImplementation on the efficient default. The high-risk tier earns a frontier reviewer.blast-radius routing
  6. 06
    Implements against the testsThe agent works inside the assembled context, with task memory that survives compaction.Mnemos
  7. 07
    Validates functionally and visuallyBuild, end-to-end and screenshot proof attached to the change.Validation Center
  8. 08
    Convenes independent reviewFive seats (architect, backend, frontend, tests, security) on models that did not write the code. One blocker caught and fixed.Council
  9. 09
    Records everythingContext, model, tests, review and outcome in the audit ledger. PR #482 opens with 8/8 checks.audit + provenance
ONE HARNESS · THREE OUTCOMES

Everything an agent needs to engineer like your best team.

CODE GRAPH · billing-svc2,418 symbols · 11 ADRs
impactpayments/refund.py → ledger/post.py → invoices/render.py
adrADR-041 idempotent money movements
dangerledger.post is append-only
teststests/payments/test_refund.py (required first)
reviewcouncil: architect · backend · tests · security · red-team
verdictAPPROVE after 1 fix
HOWlive code graph · org skills · multi-model council · functional and visual validation
ARCHITECTURE AND DEPLOYMENT

Centrally managed. Runs where you need it.

Sensitive work stays on your infrastructure. Tenant code runs in your CI, never in Srooter. Prompts are hashed and redacted for audit, never stored in plaintext.

SaaSSelf-hosted gatewayBYOKLocal / sovereign modelsSSO (OIDC · SAML)Audit exportData residencyFull VPC harness · roadmap
Read the platform-team checklist →
agents
Claude CodeCodexCursorsrooterctl
runs ondev machines · your CI · your VPC
harness
srooterSaaS or self-hosted gateway
policyallowlists · budgets · SSO · audit ledger · residency
models
frontieropensovereign / localBYOK
AVAILABLE EVERYWHERE

In your terminal, on your desktop, inside the tools you already use.

Studio brings Tasks, the Validation Center, the Council Chamber, Knowledge and Routing into one desktop app for macOS. The gateway, agent and shell run everywhere else.

Srooter Studio for macOS: the Council Chamber reviewing a change, with five independent review seats, a security blocker caught and fixed, and the code graph, validation and audit trail
srooterctlThe CLI agent and session shell. One install, session-scoped routing, no global env changes.Install →
Studio for macOSTasks, Validation Center, Council Chamber, Knowledge and Routing in one desktop app.Download for Mac →
Windows and LinuxGateway, agent and shell for every developer machine and CI runner.Download →
Inside your toolsClaude Code, Codex, Cursor and aider connect with a single endpoint change.See integrations →
PROOF

Spend frontier intelligence where it matters.

We build a real app through the harness with pinned models and score it on four axes: accuracy, speed, cost and quality. The results drive our routing defaults, and they are public.

58×Cost spread, same resultSix frontier models built the same app to the same acceptance criteria.
4Axes scoredAccuracy (Playwright), speed, cost (tokens by price), quality (screenshots).
12Providers, incl. localAnthropic, OpenAI, Google, DeepSeek, GLM, Kimi, Groq, Ollama and more.
5Independent review seatsArchitect, backend, frontend, tests, security, plus an optional red team.
Benchmark conditions, task, commit, model versions and per-run costs are published with every result.See the live benchmarks →

The tools and models will change.
Your engineering system stays yours.

Start free →Get Started
See the benchmarksExplore self-hostingRead the docs
srooter> — the engineering harness for AI coding agents