Say what should be true when the work is done. Codelit builds the smallest useful Team, reuses safe setup, and chooses the first capable run. Flow and Advanced stay available when you want the visual design tool, but they never block ordinary work.
Ask Codelit to build, edit, run, schedule, inspect, or repair the Team in ordinary language.
See one compact plan. Safe reversible setup is already complete; only real blockers need you.
Start with the least expensive capable mode and approve consequential actions before they happen.
Use Activity to inspect proof, then continue, retry, fix, or schedule from the same composer.
Specialists with roles, inputs, outputs, model preferences, escalation rules, and tool access.
Progressive instruction packs with activation rules, resources, deterministic scripts, and risk labels.
Tool, resource, prompt, roots, and sampling boundaries for real context and action systems.
Slack, GitHub, Notion, Linear, Jira, Stripe, browser workers, databases, and custom APIs.
Pick cheap models for classification, stronger models for planning, and fallbacks for reliability.
Human approval, read-only defaults, audit logs, data boundaries, and blocked irreversible actions.
Eval, sandbox, replay, approval, observability, and prompt-regression gates before rollout.
Trace history, runbooks, source ledgers, and durable state planned before the agent learns bad habits.
Reviewed files and repository excerpts indexed in the browser, with secret warnings and durable deletion controls.
Nine deterministic resilience scenarios, one-click control fixes, and redacted public scorecards with zero execution spend.
Review OpenAPI 3, MCP, and n8n mappings with credentials, expressions, scripts, and URLs discarded before replacement.
Repeat one BYOK workflow while Codelit is open, with cross-tab deduplication, one catch-up, durable controls, and one-click hosted promotion.
Plain-language definition
It is a supervised group of focused AI specialists working together on one outcome. Codelit can prepare and run the Team from one request while keeping every responsibility, input, output, tool, permission, handoff, and stop condition visible in Flow.
Describe what done looks like before opening configuration.
Scope every source and require approval before consequential writes.
Keep terminal events, decisions, evidence, costs, and artifacts in Activity.
Use Plan & Ship to create the Product Plan, Architecture, repo pack, and GitHub handoff.
What You Get
Ask Codelit is the default. It builds and changes the Team, reuses safe setup, and chooses the least expensive capable runtime. Flow keeps direct-card editing available, Advanced reveals deeper policies, and Activity keeps every run, decision, and receipt.
One composer for building, editing, connecting, running, scheduling, inspecting, and repairing.
One compact interruption only for authentication, ambiguity, meaningful cost, or an external action.
Optional Flow with direct-card editing, clear handoffs, permissions, and approval gates.
Optional Advanced controls for models, Skills, MCP, retries, scopes, evaluations, and harnesses.
Activity with terminal events, decisions, costs, evidence, artifacts, and immutable receipts.
Browser, connected-app, local, in-tab, managed, and hosted execution behind the same request.
One-click path from proven evidence into Product Plan, Architecture, repo pack, or GitHub handoff.
Starter Templates
Choose a real job and tell the Team what to accomplish. Codelit keeps the proven responsibilities and safety boundaries, then adapts only what your outcome needs.
Reads one release channel, prepares a concise owner-ready update, pauses on the exact message, posts it once, and returns Slack delivery proof.
Reviews recent deployments for one project, pauses before spending build resources, creates one preview from the latest source, and returns its deployment evidence.
Reads one repository's open issues, drafts an evidence-backed follow-up, pauses on the exact issue comment, posts it once, and returns the comment proof.
Reads one Jira project, converts a request into a scoped work item, pauses on the exact fields, creates it once, and returns the Jira issue proof.
Reads one Linear team's queue, turns incident evidence into a concise follow-up, pauses on the exact issue, creates it once, and returns Linear proof.
Reads one source page, distills a reusable decision or runbook update, pauses on the exact page content, creates a child page once, and returns Notion proof.
Gives one Browser Operator a visible goal, watches the live page, handles sign-in or MFA through takeover, verifies the result from current DOM evidence, and returns a screenshot-backed Run Receipt.
Qualifies an inbound lead against the team's real ICP, explains the score, pauses on ambiguous routing, and returns a signed handoff to the CRM or automation that sent it.
Screens a candidate against a role rubric, records evidence for every criterion, pauses before a shortlist decision, and returns a structured recruiter handoff.
Classifies incoming document text, checks routing and duplicate signals against a connected Jira project, pauses on sensitive cases, and returns an auditable work-queue handoff.
Turns one approved Notion source into channel-specific drafts, checks claims and voice, pauses before publication, and returns a ready-to-schedule content package.
Turn an inbox request into a researched draft, calendar handoff, and updated operating record. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Research a work request, prepare a draft, schedule the handoff, and publish the approved update to Teams. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Find a customer ticket, prepare a grounded resolution, and add the approved comment or status update. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Find a conversation and post one approved admin reply with a durable receipt. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Verify a payment, compare it with policy, and issue one approved full or partial refund. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Inspect an order and apply one approved operational tag for fulfillment or resolution follow-up. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Review a contact and move its lead status only after the evidence and owner approval are present. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Review a lead and create one approved follow-up task without broad SOQL or field access. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Read the highest-impact issues and apply one approved status change after diagnosis. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Review saved insights and attach one approved operational annotation to the product timeline. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Review active monitors and post one approved incident event to the operational timeline. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Review active incidents and acknowledge one selected incident after ownership is clear. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Read an RLS-scoped table and insert one schema-reviewed operational record. The team reads one bounded scope, shows the exact change, executes it once after approval, and returns provider evidence.
Turns one scoped release request into a reviewed branch artifact, preview deployment, browser-verified result, pull request, production promotion, and Slack proof using the accounts the team already owns.
Correlates Sentry, Datadog, PagerDuty, and deployment evidence, proposes one bounded mitigation, pauses on the exact actions, and leaves status proof for responders.
Reads one support ticket, order, and payment, evaluates the request, issues only the approved bounded refund, updates the customer ticket, and leaves financial and support proof.
Combines CRM context with bounded browser research, explains the qualification decision, creates an unsent email draft, and updates the lead after approval.
Turns a Slack request and an approved Google document into a cited answer, appends only the reviewed update, responds in-channel, and preserves both provider receipts.
A browser-only team that reviews explicitly selected local documents, separates evidence from assumptions, and produces an owner-ready brief without connected apps or managed infrastructure.
A Slack-native engineering agent that receives operational requests, gathers context from tickets and repos, routes work to specialist agents, and drafts auditable responses before anything risky happens.
A code review workflow that watches pull requests, assigns specialist reviewers, checks implementation risk, and posts concise review notes with links to evidence.
A controlled workflow for small teams that want an internal agent to answer operational questions, create tickets, inspect connected apps, and prepare approved actions across the company stack.
A guarded browser operator for internal tools: it can open approved apps, complete repetitive workflows, verify the result, and stop at human approval checkpoints before sensitive writes.
A Devin-style engineering workflow that turns a scoped ticket into a branch, implementation plan, code changes, test run, and pull request draft with review evidence.
A support workflow that reads product docs, checks account and billing state, drafts empathetic replies, and escalates sensitive requests with source-linked evidence.
A meta-agent workflow that generates red-team cases, runs regression suites, scores agent behavior, and blocks production rollout when a workflow violates its safety contract.
A research workflow that gathers competitor signals, checks source quality, extracts positioning patterns, and writes briefs that separate facts from interpretation.
A security workflow that watches alerts, gathers evidence from code and runtime systems, ranks blast radius, and prepares a human-approved remediation plan before any production action.
A finance-aware workflow that inspects subscriptions, invoices, usage limits, and customer requests, then drafts revenue-safe actions for approval.
A launch workflow that coordinates release notes, docs, changelog updates, social copy, customer comms, and post-launch monitoring from one evidence-backed plan.
A data-quality workflow that watches analytics schema changes, validates event health, explains metric drift, and opens owner-ready fixes when instrumentation breaks.
A Claude Code-style workflow for scoped engineering maintenance: it reads issues, loads repo context, plans a narrow patch, runs checks, and opens a reviewable PR with evidence instead of pretending the merge is done.
A code-first support agent built around specialist ownership, function tools, MCP resources, guardrails, traces, and eval loops so customer answers can move fast without leaking account state.
A LangGraph-style incident workflow that keeps explicit state, branches on severity, retries evidence collection, pauses for human approval, and replays traces after the incident is closed.
An n8n-centered workflow that turns a business automation request into a tested app workflow with conditional logic, human-in-the-loop checks, cost controls, and failed-run repair notes.
A CrewAI-style research desk with role-based agents for source discovery, analysis, critique, and publishing, built for briefs that need evidence instead of vibes.
Flow keeps direct editing available; Advanced holds models, retries, scopes, and policies.
Flow guideCodelit starts with the least expensive capable runtime; Run options remain available.
Run mode guideTurn accepted workflow evidence into reviewed product and engineering artifacts.
Plan & ShipCommon questions
An AI Agent Team workflow coordinates specialist agents, tools, handoffs, decisions, and human approvals around one measurable job. It makes the operating flow visible before a model or integration is allowed to act.
Yes. Start with one message in Ask Codelit or a proven template. Codelit proposes the smallest useful Team, handles safe setup, and asks only when access, cost, ambiguity, or an external action needs you. Open Flow or Advanced only when you want to inspect a decision.
A team can be dry-run for free, run in an open browser tab with a user key, use a local workspace, or use Codelit's managed Live and hosted runtimes when the workflow needs connected actions, browser work, or unattended schedules.
Agent Teams are the primary build-and-run workflow. Plan & Ship turns a proven team and its evidence into a reviewed Product Plan, system Architecture, repo pack, and GitHub handoff.
Build, edit, Sample, and run eligible work with your own key for free. Pro adds managed models, browser work, automatic recovery, delivery, and unattended schedules without making expert controls mandatory.
Try agent