All functions

AI agent teams

Build a team of AI agents you can trust.

Give every specialist a clear role, model, memory, tools, project access, and budget—then bring that agent into the work.

See what agents can do

CanDoItAll technical agents organized into reusable specialist teams — Teams **Meet the whole agent team.** Organize reusable specialists by role and team so people can quickly find the right help.

Delivery Manager agent adding sequenced tasks while floating over a CanDoItAll project structure — In projects **Bring a specialist into the project.** A floating agent can inspect the project structure, propose work, and keep its actions visible in context.

CanDoItAll process preflight assigning an agent team and exact AI agents to required process roles — Process roles **Assign exact agents to every process role.** Choose a prepared team, resolve each role before launch, and keep the staffing decision with the process that will run.

What it covers

Everything an agent needs to join the work.

A clear role, chosen model, memory, capabilities, and access rules travel with the agent wherever it appears.

Technical identities and teams

Create, clone, import, export, group, and template agents with stable role and instruction definitions.
Provider and model binding

Select provider profiles and models per agent, with runtime setup verification before use.
Memory and modalities

Configure memory behavior, image generation, and provider-supported voices independently for each agent.
Granular access

Separate permissions cover project reads, project creation, subprojects, task changes, broader structure, processes, workspace files, storage catalogs, and secret references.
Unified capabilities

Attach tools, skills, and MCP servers from one searchable catalogue while keeping the assigned surface inspectable.
Contextual collaboration

Open focused or floating chats from projects, Gantt, workflows, processes, calendars, and scheduler surfaces.
HR-assisted team design

Use the built-in HR agent to identify the right specialist and help create or configure agents for the work.
Inspectable execution

Review approvals, artifacts, checkpoints, tool receipts, logs, usage, known cost, and recent failures.
Business-role binding

Bind a CRM / HR AI party to a technical agent so organizational roles resolve to a governed runtime identity.

Agent settings

Set the boundary once. Keep it easy to inspect.

Start with identity and runtime. Add only the memory, tools, project access, secrets, process access, and voice that the role needs.

CanDoItAll agent Identity tab with name, role, instructions, and visual identity settings — Identity **Give each agent a durable identity.** Define its name, role, purpose, instructions, visual identity, and reusable template boundary.

CanDoItAll agent Runtime tab with provider, model, workload, and policy settings — Runtime **Bind the execution runtime explicitly.** Choose the provider, model, workload, chat-history policy, diagnostics, and runtime behavior this agent should use.

CanDoItAll agent Memory tab with provider, invocation, tool, and failure settings — Memory **Control how memory participates in the work.** Select the memory provider, invocation mode, exposed memory tools, and failure behavior for this agent.

CanDoItAll agent Images tab with image provider, model, and project asset settings — Images **Give image generation an explicit policy.** Enable image generation per agent, bind a provider and model, and decide whether outputs are saved to project assets.

CanDoItAll agent Project Structure Access tab with granular project permissions — Project structure access **Grant project permissions separately.** Read, create, restructure, subproject, and task-change permissions remain visible and independently controlled.

CanDoItAll agent Workspace Tools tab with file, workspace, and storage access settings — Workspace Tools **Bound file and workspace access.** Execution tools, workspace roots, storage catalogs, and file access are granted deliberately instead of inherited implicitly.

CanDoItAll agent Secrets tab with explicitly allowed secret references — Secrets **Allow only the secrets an agent needs.** Grant explicit secret references without placing credential values inside the agent definition or its instructions.

CanDoItAll agent Process Access tab with read, write, and process allowlist settings — Process Access **Scope access to process definitions and runs.** Separate read and write permissions, then allow all processes or maintain a definition-level allowlist.

CanDoItAll agent Capabilities tab combining tools, skills, and MCP servers — Capabilities **Attach catalogued capabilities in one place.** Search and attach tools, skills, and MCP servers through a single inspectable capability catalogue.

CanDoItAll agent Voice tab with audio access and provider voice selection — Voice **Give each agent its own voice.** Enable audio mode and select a provider-supported voice per agent; voice controls are available across CanDoItAll chats.

Use the arrows or choose a setting to inspect the current interface. Selecting a step stops the automatic tour and keeps that screen in place.

Agent and project analytics

Trace usage, outcomes, and cost back to the work.

CanDoItAll captures activity from chats, workflows, and processes and, whenever project context is available, attributes that activity to the project that authorized it.

Compare providers and agents, review project history against the remaining plan, search process cost over time, and ask a process manager about a specific run.

The deeper Economy module is planned for later. Today’s operational records keep the available cost trail—and any unknown or estimated values—visible.

CanDoItAll agents dashboard with total expense, token use, sessions, provider distribution, most-used agents, and failed runs

Agent overview

See total agent usage and spend at a glance.

Total expense, token use, sessions, provider ratio, most-used agents, and failed runs reveal the operating mix behind AI work.

CanDoItAll project manager summary with historical and planned costs, cost ratio, recent activity, and expense trend

Project costs

Connect actual spend to the remaining project plan.

Known and estimated history sit beside planned cost, category ratios, recent activity, and trends for the selected project scope.

CanDoItAll project Gantt with task, dependency, effort, assignment, date, and expected-cost information

Plan exposure

Put expected cost beside schedule and effort.

The Gantt view keeps task dependencies, delivery and pure effort, assignments, dates, and expected cost in one planning scope.

CanDoItAll manager summary activity dialog with filters, aggregate costs, durations, and a detailed list of agent runs

Activity costs

Inspect the individual runs behind a total.

Filter agents, workflows, or processes by outcome to review activity time, known cost, recorded estimates, and status.

CanDoItAll process history graphs with cost, tokens, runtime activity, manager signals, and tool usage

Process trends

Search process cost and telemetry over time.

History filters reveal cost, token, runtime, manager-signal, and tool-usage patterns across process runs.

CanDoItAll process manager chat explaining the cost and token breakdown of a selected process run

Ask a run

Talk to the process manager about one run.

A run-scoped manager can explain actual cost, token breakdown, outcome, and execution context without leaving process history.

Capture now

Retain provider, model, tokens, duration, status, known cost, estimates, and run context for conversations and automated work.

Attribute to projects

Connect chats, agents, workflows, processes, people, and planned work to the project scope whenever the relationship is available.

Investigate the total

Move from summary ratios and trends to individual activities, then ask the process manager about the exact run that needs attention.

How it connects

Microsoft Agent Framework is a strong foundation for agents.

Microsoft Agent Framework provides the core runtime for agent execution and orchestration. CanDoItAll adds the adoption layer around it: reusable agent identities, provider profiles, voice, permissions, project assignments, budgets, cost tracking, and governance.

Browse Microsoft Agent Framework on GitHub Next: CRM / HR

Build a team of AI agents you can trust.

Everything an agent needs to join the work.

Technical identities and teams

Provider and model binding

Memory and modalities

Granular access

Unified capabilities

Contextual collaboration

HR-assisted team design

Inspectable execution

Business-role binding