For entrepreneurs

Your AI team,
in one workspace.

Give each agent a role. Keep the context.
See where the work stands.

Private release · Bring your own API key

Explore the five views

A place for the work

One workspace.
Five ways to stay in touch.

Illustrative examples, not live agent activity.
Availability is noted on each card.

01 / InstructionsIllustration

One tree. A clear role for each agent.

  • Company context
    • Product principles
    • Customer knowledge
  • Engineer’s role
    • Review before shipping
    • Report what changed

Organize instruction and knowledge pages, assign recipients and keep the change history.

Management available
02 / HandoffsIllustration

Keep the next step in context.

Researcher→Writer
REQUESTED NEXT ACTION

Turn the findings into a short product brief.

Research notesSource links

Carry task lineage, context and artifact references to the next agent. Live agent-to-agent execution still needs validation.

Lifecycle implemented · Live validation pending
03 / AgentsIllustration

Your agents, one view.

General assistantBoard
ResearcherInstructions
EngineerMemory

Choose whose work to see. Each agent has its own role; the Board keeps its to-dos and Brief close together.

Management available
04 / Memory & filesIllustration

Useful knowledge, beyond a chat.

MEMORY

Keep the product brief short and include the source behind each claim.

Source review needed
Product brief.mdRevision 3
Research notes.mdRevision 2

Inspect saved memory and file records. Track sources and review changes. Agent writes to Tama memory and file records passed live acceptance.

Records and reviews available
05 / People & permissionsIllustration

Keep people in the decision loop.

DECISION NEEDED

The source changed. Does this instruction still apply?

Review the evidenceRecord a decision

Permissions belong to the task and its scope.

Review task approvals and knowledge changes. Tama can review knowledge with permission; proposed content changes await your decision. Team invitations remain planned.

Manual review available · Team tools planned
Read the workspace guide →

Give each agent a clear role.

See where the work stands.

Keep shared knowledge up to date.

Be easy and efficient

The work should stay understandable as it moves between people, agents and tools. Tama brings instructions, context and decisions into one place, with a clear record of what changed.

Pipeline combines permitted Tama analysis with human decisions. General-purpose scheduled agent tasks remain work ahead.

What is here.
What comes next.
What stays yours.

NOW / ORGANIZE

A workspace for the whole picture.

Agents, instruction pages, knowledge versions and a review Pipeline. Keep context and decisions together.

NOW / BYOK

Your connection. Clear costs.

Connect a provider API key you control. Provider usage is billed to its account. Tama service plans and limits have not been announced.

NOW / EXECUTION

Real work, verified end to end.

BYOK chat, same-session Knock and saved results passed live acceptance. Live agent-to-agent handoffs remain a separate validation item. Illustrations are not live activity.

DESIGN PRINCIPLE / CONTINUITY

Keep a record. Show uncertainty.

Tama stores its own work records. An uncertain run needs review, not a silent retry. Provider-session and storage lifetimes remain separate.

Start with one agent

A role. Some context.
A place to begin.

Read the guide →
  1. 01

    Sign in

    Use your authorized Google account or workspace token.

  2. 02

    Choose an agent

    Start with Tama or add an agent with a clear role.

  3. 03

    Set its instructions

    Organize the pages it should use.

  4. 04

    Prepare the work

    Write a task and Knock, or send a message in Code.

  5. 05

    Review progress

    Read the Brief and review results and decisions.

Connect your API key in Settings. Chat and Knock need a ready execution service.

Questions

The things to know
before you start.

All questions →
What is an agent in Tama?

A working role with its own identity, instructions, memory and assigned work. Tama manages those records separately from the model provider's execution session.

Who pays for model usage?

The current private release uses bring-your-own-key (BYOK). The workspace owner connects a provider API key and pays that provider for its usage. Tama service pricing and public plans have not been announced.

Do I need a Claude or ChatGPT subscription?

No personal subscription is connected by this release. A consumer subscription does not replace a provider API key. Access to existing Codex App conversations or subscription allowances is not offered here.

What does Knock do?

Knock queues a request for an agent to review updated instructions, report progress and handle its to-dos. New work waits for the current task. Chat and Knock are available in the private release with a configured API connection and execution service.

What stays between conversations?

Tama stores workspace instructions, knowledge, memory and work records. Provider conversations, files and execution environments have separate lifecycles. This is not a promise of permanent storage or unlimited recovery.

What does outdated mean?

Knowledge may need review because a source changed or someone marked text as outdated. That is different from an instruction update that has not yet been delivered to an agent. Neither a color nor delivery status proves the model understood it.

Tama robot

A home for your agents

Keep the work together.

Instructions, knowledge and decisions.
Room for your next idea.