EvolutionID · IDfunction Agent · v1.0

Coding agents,
in your browser.

Access Claude Code, DeepAgents, Codex, and OpenCode through the browser. Switch harnesses and models on any repo in seconds — no local setup, no configuration, no wasted time.

Coding harnesses3 LIVE · 1 SOON
Claude Code DeepAgents Codex OpenCode soon
Switch harnesses on any repo without local setup. State carries across — pick the right tool per task.
Models4 PROVIDERS
ClaudeGPTGeminiOpen-source via OpenRouter
Bring your own keys or route through OpenRouter for hundreds of open-source models — frontier or specialised, you choose.
Sandbox
MicroSandbox · KVM-isolated
CI/CD
GitHub Actions · auto-wired
Deploy
Cloud · SSH soon
Monitor
Stack health · live probes
SOC 2-aligned controls AES-256 at rest SSO · Okta · Azure AD Enterprise support
agent.evolutionid.com / workspaces / acme-hq
Agent ready
Workspaces
acme-hq
northwind-eu
tessera-tower
Files
rbac.policies.yaml
badge_workflow.ts
pacs/genetec.ts
docker-compose.yml
migrations/0014.sql
acme-hq / claude-code
microVM · ready
Sonnet 4.5
DU
Set up RBAC for Acme HQ. Three roles — Employee, Contractor, Visitor. Contractors need badge expiry after 90 days and approval from facilities before issuance.
/
Got it. Working inside the sandbox — I'll add the role definitions, the contractor approval workflow, and a migration to seed the policies.
edit_files+3 files✓ 1.4s
+ rbac/roles.ts — Employee · Contractor · Visitor
+ workflows/badge.contractor.ts — expires_after: 90d
+ migrations/0014_seed_acme_roles.sql
/
Three files staged. Want me to wire the Genetec PACS adapter next so badges
> Yes — pull credentials from the encrypted vault
02 · The thesisWhy multi · why now

One workspace. Many tools.

No single harness, model, or runtime wins every task. IDfunction Agent is built on four axes of optionality — so the right choice is always a switch away, and the work survives the swap.

01 / 04
Multi-harness
because no single harness is best at all tasks

Claude Code excels at code edits and long-context refactors. DeepAgents shines at structured tool-using planning. Codex is fast for surface tweaks. Switch per task; state carries across the session.

4 harnesses · live state hand-off
02 / 04
Multi-provider
because cost/quality tradeoffs depend on task

Frontier Claude for hard architecture work, Haiku for boilerplate, GPT for one-shot generation, open-source via OpenRouter for high-volume runs. Choose the cost curve per project, not per platform.

4 providers · per-workspace routing
03 / 04
Sandboxed
because security is not a feature flag

Every workspace runs in a MicroSandbox-managed microVM with KVM isolation. No host filesystem access, isolated network, configurable CPU and memory. The agent can break things — just not your things.

KVM microVM · per workspace
04 / 04
Organization-aware
because each team has different policies

Roles, model access, harness allowlists, audit retention, and billing scoped to your organization. Invite a team, set the policy, ship.

Orgs · roles · audit · billing
03 · ArchitectureTwo systems, one workflow

One platform. Two systems. Zero boilerplate.

IDfunction Agent separates the tooling from the output. The AI Harness is the cloud workspace where you talk to the agent. The PIAM Product is the per-customer system that ships to the customer's environment.

AI Harness · Cloud
Where you work
agent.evolutionid.com
Browser chatFile treeGit panelmicroVM sandboxCanonical sessions
Agent generates
PIAM Product · On-prem
What ships
idfunction-{client}-{project}
Docker ComposePostgreSQLVendor adaptersReverse proxyHealth probes
04 · Canonical sessionsOne transcript, every harness

A single session model. Across every harness.

Each harness streams its own raw events; our canonical session layer normalises them into a typed, ordered transcript — keyed to your workspace, independent of which harness or model produced it. Replay it. Fork it. Audit it. Migrate it.

Claude Code · raw
tool_use(edit_file) path: src/auth.ts +12 -3 lines result: applied
DeepAgents · raw
{"action": "write_file", "args": {"path": "...", "content": "..."}, "thread": "th_8a2"}
Codex · raw
[function_call] name: file.write args: {path, body} call_id: c_9f3a
Normalised · sessions/sess_a1b2c3
Canonical event streamREPLAYABLEFORKABLEAUDITABLE
event 21
user.message "add zod validation to the login route"
event 22
agent.tool_call read_file(src/auth.ts) → ok
event 23
agent.tool_call edit_file(src/auth.ts) → +12 -3
event 24
agent.tool_call shell(npm test) → 1 failing
event 25
agent.message "Tests fail on edge case. Want me to handle it or revert?"
event 26 ↳ fork
branch.created sess_a1b2c3.f01 · model=gpt-4o · differential eval
01 / 09

Compliance unit

For regulated customers asking "what did your agent do for us last quarter?", a canonical transcript is the only defensible answer.

02 / 09

Replayable evals at scale

Every production session becomes a labeled test case. Re-run against a new model, diff outcomes, gate promotions.

03 / 09

Forking · what-if

Pause an agent at turn N, fork with a different tool result, see what happens. The killer feature for debugging failure modes.

04 / 09

Time-travel debugging

For support: "agent did X for customer Y, replay from event 23 with verbose logging." Reproducibility from the audit trail.

05 / 09

Sessions as an API

External tools — Slack, ticketing, dashboards — read agent state without reaching into a specific harness or runtime.

06 / 09

Migration insurance

When you add a fourth harness or swap one out, your session investments survive. The canonical layer is forward-compatible.

07 / 09

Differential evaluation

Same session, two models, run side by side. A pattern only practical when the session is independent of the runtime.

08 / 09

Cross-session memory

Workspace-scoped memory survives session boundaries and harness swaps — long-context grounding that travels with the project.

09 / 09

Multi-agent orchestration

Orchestrator + worker sub-agents writing into one session, with explicit ordering. Structured collaboration instead of ad-hoc sub-process trees.

05 · Harness configurationFully tunable · per workspace

Configure every harness. Per workspace.

Most platforms ship harnesses with a fixed personality. We don't. Pin system prompts, scope tool allowlists, set token and iteration ceilings, route to the right model, control approval gates — independently for each workspace, for each harness. Standardise per customer; experiment per branch.

"Same harness, different rules. Acme HQ runs Claude Code with strict approval gates and a small tool surface; the internal scratch workspace runs the same harness wide open."

System prompts

Pin a custom system prompt per harness, per workspace. Append a customer-specific preamble without modifying the underlying CLI.

Tool allowlists

Whitelist exactly which tools the harness can call — file read, shell, web fetch, package install. Deny by default for regulated workspaces.

Model & provider routing

Pick the model per harness per workspace. Frontier Claude for hard work, Haiku for the cheap path — and an override for client-billed projects.

Approval gates & limits

Require human approval for shell commands, deploys, or destructive ops. Cap iterations, token spend, and runtime per session.

Env vars & secrets

Inject workspace-scoped environment variables into the sandbox. Credentials live in the AES-256 vault; the harness sees them via the runtime.

Claude Code
DeepAgents
Codex
OpenCode
Workspace · acme-hq SAVED · 2 MIN AGO
Model · routing
Primary modelclaude-sonnet-4.5Override
Fallbackclaude-haiku-4.5Org default
Temperature0.2Slider
System prompt · workspace preamble
You are working in the Acme HQ workspace. Follow Acme's coding standards in /docs/CODING.md. Never modify /migrations/published/*. Prefer Postgres over Redis for new state…
Tools · 5 enabled · 2 blocked
read_fileedit_fileshell (gated)git_*http_gethttp_postinstall_package
Approvals & limits
Auto-approve edits
Shell needs approval
Network access
Max iterations40per turn
Token budget200K · sessionHard cap
06 · How it worksFrom idea to deploy in four steps

From idea to deployment. In four steps.

A guided pipeline that compresses a fortnight of manual setup into an afternoon of conversation.

STEP 01
Create
A 4-step wizard: project info, harness & model, server config, review. Fork a template or attach to an existing repo.
/workspaces/new · ~3 min
STEP 02
Provision
Agent creates a dedicated GitHub repo, configures CI/CD, provisions a server, and spins up a MicroSandbox microVM — tracked in a live progress view.
GitHub · DigitalOcean · GHCR
STEP 03
Converse
Open the workspace, pick your harness, describe what you need. Switch models mid-task; switch harnesses at any time. The canonical session preserves the thread.
3 harnesses · 4 providers · 1 session
STEP 04
Deploy
Three-state git tracking — Unsaved, Unpublished, Published. One-click commit and push; CI builds the Docker image and ships it to GHCR.
Save · Publish · Preview
07 · FeaturesEight capabilities, one workspace

Everything you need to ship a customer project.

A workspace built around the way developers actually work — chat, files, git, shell — with sandbox isolation, multi-harness session continuity, and team-aware controls wired in.

01 / 08

Harness-agnostic sessions

Switch from Claude Code to DeepAgents mid-task. Your session, tool history, and pending state carry over. No conversation loss, no context rebuild.

3 live · canonical event log
02 / 08

KVM-isolated microVMs

Each workspace runs in its own MicroSandbox microVM. Hardware isolation, no host filesystem access, configurable CPU and memory.

MicroSandbox · KVM · OCI image
03 / 08

Bring your own model

Switch between Anthropic, OpenAI, Google, and OpenRouter per project. Per-workspace overrides for client-billed work.

4 providers · per-workspace routing
04 / 08

Session forking

Pause at any event, fork with a different model, prompt, or tool response. Run the branches side by side and compare outcomes.

Branch · differential eval
05 / 08

Time-travel replay

Replay any session from event N with new verbose logging — for support, debugging, or regression testing against a newer model.

Replay · verbose mode
06 / 08

Teams & organizations

Invite your team, scope models and harnesses by role, aggregate billing, enforce policies. Built for shared workspaces from day one.

Org · roles · audit · billing
07 / 08

Resilient sessions

Page refresh, tab close, network blip — the runtime stays alive for 5 minutes. Reconnect and the hello frame restores state and history.

5 min grace · question replay
08 / 08

Production Docker output

Multi-stage builds, health checks, reverse proxy. Semver-tagged in GHCR, exportable as .tar.gz for air-gapped deployments.

On-prem · air-gapped · semver
08 · The workspaceYour AI-powered development environment

Your AI-powered development environment.

Real-time chat with streaming. A file tree with git-native workflow. Live agent state. Harness and model switching. Built for the way teams actually ship.

Filesmain
● Saved○ Unpub 3
▾ acme-hq
▾ rbac
policies.yaml
roles.ts
▾ workflows
badge.contractor.ts
▾ pacs
genetec.ts
▾ migrations
0014_seed.sql
docker-compose.yml
README.md
acme-hq/evolutionid/acme-hq-piam
AGENT IDLESONNET 4.5
DU
Wire the Genetec PACS adapter and run a deploy preview on staging.
/
Reading credentials from the project vault, then scaffolding the adapter.
shellnpm i @genetec/sdk✓ 2.1s
+ pacs/genetec.ts — adapter scaffolded+ credentials read from vault://acme-hq/genetec+ health probe registered at /healthz/pacs
git_commit"wire genetec adapter"RUNNING…
staged → 4 filescommit → 7a3b21fpushing to origin
>Send a message to the agent…SEND
1

File tree

Three-state git tracking — Unsaved, Unpublished, Published — on every file.

2

Repo link

Each workspace has its own GitHub repository under your org.

3

Agent state

Live state machine — Thinking · Executing · Waiting · Idle.

4

Harness badge

Switch between Claude Code, DeepAgents, and Codex. Canonical session carries over.

5

Tool calls

Every tool execution named, scoped, and timed inline. Shell, edits, git, deploy.

6

Composer

Streamed prompts; input gates on agent state; multi-choice overlay for selectors.

09 · Teams & organizationsOnboard the whole team

Built for teams. Not just individuals.

Agentic coding is a team sport. IDfunction Agent gives organizations shared workspaces, role-based access, aggregated audit, and a single billing surface — so you can onboard your team to agents on the harnesses they actually want to use.

Organization-first model

Workspaces, repos, and sessions belong to your organization — not individual users. Bring your whole team into one shared surface.

Role-scoped access

Admin, Developer, Viewer. Scope harness allowlists, model access, and workspace permissions per role.

Centralised credentials

One organization key set, used by every member. AES-256-GCM encrypted; rotate from a single page.

Aggregated audit & billing

Every session — across members, harnesses, and models — rolled up into one org-level audit trail and one bill.

Policy enforcement

Require approval gates for production deploys. Block specific tools or models per workspace. Apply org-wide retention.

Onboarding flow

Invite teammates by email; they pick up the harness, the sandbox, the canonical session — already wired for agentic coding.

evolutionid.com · Members14 seats · 11 used
MemberRoleHarness accessLast session
DUDominik UlrichAdminAll · 32 min ago
MSMira SchenkDeveloperCC · DA14 min ago
JKJamil KanaanDeveloperCC · Codex1 h ago
RPRina PetrovićViewerRead-onlyYesterday
TSTheo SondergaardDeveloperDA only3 days ago
10 · SecurityEnterprise-grade by default

Hardware isolation. Auditable trails.

Designed for environments where access control isn't optional and audit trails aren't negotiable. Every credential is encrypted, every action is logged, every workspace runs in its own VM.

KVM microVM

Each workspace runs in a hardware-isolated MicroSandbox VM. No host filesystem access, isolated network.

AES-256-GCM

Every API key and credential encrypted at rest with authenticated encryption. Same key in frontend and runtime.

OIDC / SSO

NextAuth v5 with Azure AD, Okta, and custom identity providers. JWT sessions stored in Redis.

SOC 2 · ISO 27001

Canonical session is the audit trail. Structured logs for every workspace and agent action, with end-to-end correlation IDs.

Rate-limited & sanitised

Per-user rate limiting via Redis. Path-traversal protection and shell-injection blocking by default.

11 · By the numbersWhat ships in v1.0
3+
Live harnesses
Claude Code · DeepAgents · Codex
4
Model providers
Anthropic · OpenAI · Google · OpenRouter
1
Canonical session
One transcript across every harness and model
AES-256
Encryption at rest
GCM authenticated · all credentials
Ready to ship faster

Onboard your team to agentic coding.

Open the workspace, invite your teammates, pick a harness. Every session, every harness, every model — one canonical record.