Your agent. Your laptop. Now reachable from Slack.

Ask from Slack and the coding agent already running on your own computer picks it up — your login, your files, your approved projects. It works privately until you decide the answer is worth sharing.

Works with the agent you already run — Claude, Codex, Cursor, Devin, Gemini, Kilo, or OpenCode.

  1. Start privately

    A channel question starts private work.

    Mention Symma in the Slack thread that needs help. It opens one top-level DM with the context; every later turn stays in that DM’s thread. A direct DM works too, but stays private because there is no source thread to return to.

  2. See the scope

    It says what it is touching before it starts.

    Symma names the project it opened and the permission level it is using, then gets to work. Attach a screenshot or a file and it reads that too. You can change the permission level for the next message. If your agent offers model choices, you can change that too.

  3. Watch the work

    One message that keeps telling you where it is.

    Instead of a blank wait, the reply updates itself with whatever your agent is doing right now — reading a file, running a test. When the answer lands, the status line goes away.

  4. Keep going

    The DM thread holds the conversation.

    Ask another question in the same thread. Symma picks up the agent session when it can and restores the conversation when it cannot. It keeps one turn moving at a time so the session never splits.

  5. Review before sharing

    Send the conclusion, not the working notes.

    Refine the answer in private. Nothing returns to the channel until you confirm exactly what should be shared.

Start privately

Scroll to follow the route

One person · one route

It works like your agent because it is your agent.

One Slack app can help a whole workspace. Behind it, every person reaches only the machine, login, projects, and conversations paired to them.

Your computer

The agent runs where you already use it, not on a shared bot machine.

personal

Your approved project

Slack can choose only from the project folders offered by your companion.

allowlisted

Your access level

Read-only and write-capable modes are real permission choices from the agent.

explicit

Your model

Symma offers the models your own agent advertises, and nothing else. Whatever you pick applies from your next message.

capability-led
Current setup

Connect the machine you already work on.

A workspace admin connects Symma once. Each person pairs the computer where their agent is already signed in.

  1. 1
    Ask Slack for a pairing code

    Run /connect in any Slack channel where Symma is present.

  2. 2
    Pair your computer

    Run the one-time command Slack gives you on the computer where your coding agent is already signed in.

  3. 3
    Start a private conversation

    Symma finds your available agent and the project folders you approved.

Who can use this today: people whose workspace admin directly installs Symma’s internal Socket Mode app, and who already run a coding agent on their own machine. There is no public sign-up yet, and Symma is not in the Slack Marketplace.

Clear boundaries

The conversation travels. Your login does not.

Symma relays the work to the correct person’s machine. The hosted service can see the content it relays. Provider credentials and the project workspace remain on your computer; anything your agent includes in the relayed work is visible to the service.

  • Work away from the channel Investigate and refine in a DM thread that belongs to you.
  • Owner-scoped Your Slack identity cannot fall through to someone else’s machine.
  • Your machine stays yours Your companion connects out; your provider login and project remain local.
  • Publish with intent Share the verified conclusion back only when it says what you need.
An alliance, not an autopilot

Keep the agent close. Keep the final word.

Symma is open source. Follow the work on GitHub and see how a personal agent can fit into the conversations your team already has.