> ## Documentation Index > Fetch the complete documentation index at: https://docs.gravitygtm.com/llms.txt > Use this file to discover all available pages before exploring further. # The Slack assistant: research, edits, and standing feedback in chat > What the conversational Slack assistant can research, write, and remember today — DM it, @mention it, or reply in a thread it's already part of. Once the [Slack app is installed](/agents-lite/getting-started#install-the-slack-app), your team has a live conversation with your fleet, not just a stream of digests and approval cards. This page covers what that conversation can actually do today — it's grown well past research-only since it first shipped, and it keeps growing. It still can't read your Gmail, Sheets, Drive, or Calendar directly as part of a live conversation — those stay automated-agent-only, and the sheet tools below can change an existing row but never read one back. Within that boundary, though, this is a real, growing set of tools, not just a chat wrapper around a search box. ## How to reach it * **Direct message** it, one-on-one. * **@mention** it in any channel it's a member of — a fresh top-level mention gets a little recent-channel context automatically, so it isn't answering with zero awareness of what was just discussed. * **Reply in a thread** it's already part of — under its own message, or under an approval card it posted. Any of the three carries the recent back-and-forth forward, so a follow-up question doesn't need to repeat context you already gave it. ## What it can do ### Research, on demand Company and people lookups, LinkedIn profile/company research, finding a verified email address for a contact, and reading recent activity across your own Slack channels to help answer what it's asked — all free, all read-only. ### Update your pipeline and investor sheets | Tool | What it does | | ---------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `update_sheet` | Updates one or more fields on an **existing** pipeline or investor row, matched by company name. It never creates a new row — a one-shot chat message is a riskier way to originate a brand-new record than to correct an existing one, so that stays with the automated agents. | | `archive_row` | Archives a row off your active pipeline or investor sheet, with a required reason. Reversible by a human looking at the archive, same as an agent-initiated archive. | | `merge_entities` | Merges two duplicate rows — you name the one to keep and the one to fold in, plus a required reason. | Each of these confirms back in the same conversation once it's done — it's not a silent background action. ### Tune your configuration The same [config tools](/agents-lite/self-serve-config#config-tools-for-an-already-live-deployment) your Claude.ai/ChatGPT connector uses are also available in Slack — `get_config`, `get_capabilities`, `update_config`, and `revert_last_change` — with one difference that makes chat the *safer* of the two surfaces for a real change: asking for an edit here doesn't write anything by itself. It posts a real **Approve / Reject** card describing exactly what would change, and nothing happens until someone on your team taps Approve. Reject (or just not tapping) leaves your configuration untouched. ### Generate contracts and proposals `generate_contract` re-renders an engagement you've already set up, and `author_contract` builds a brand-new one from scratch — gathered conversationally (deal terms, pricing, scope), optionally filling in your own contract template if you've shared one. Both deliver the finished document back to you in Slack; neither invents a figure or date you haven't actually given it. ### Ask for something that isn't configurable yet If you ask for a setting that doesn't exist yet, the assistant doesn't just say no — `request_config_change` logs the actual request and sends it to Gravity as a real signal, not a dead end. You'll still get a plain "that's not available yet, but I've flagged it" in the moment. ### Shape future drafts with standing feedback Two ways to get an ongoing correction into the system, not just a one-off edit to a single draft: * **Say it any time.** Tell the assistant something like "stop pitching the enterprise vertical, it isn't converting" or "the tone should be more direct" — `record_content_feedback` folds it straight into how future content gets written. There's nothing to confirm back beyond an acknowledgment that it was received. * **Answer the weekly check-in.** Once a week, whoever owns a draft mailbox gets a short DM asking for any standing feedback on recent outreach. A reply there feeds the exact same place the "say it any time" path does. Both are separate from editing one draft in place (tapping **Edit** on an approval card, below) — those two channels are for a *pattern*, not a single message. ## Approval cards and reactions The same cards the fleet already posts for an email or a LinkedIn-post draft are interactive, not just informational — tap **Approve**, **Edit**, or **Cancel** right on the card instead of acting on it elsewhere. A 📤 reaction on a parked, no-email contact promotes it straight into your active pipeline. See [Getting started](/agents-lite/getting-started#install-the-slack-app) for the install step these ride on top of. ## The Home tab Each teammate gets a Home tab (the app's own tab in Slack, not a channel) showing which modules are currently active for your fleet — a quick "what's actually turned on" view with no digging through settings. ## Who can use it Only your own team — the same people list your fleet already recognizes for drafts and approvals. A message from someone outside that list is politely declined, not silently ignored and not acted on. ## Next steps * [Self-serve configuration](/agents-lite/self-serve-config) — the same config tools described above, in full, plus the connector-based surface they're shared with * [Getting started](/agents-lite/getting-started#install-the-slack-app) — installing the Slack app in the first place * [Security & credentials](/agents-lite/security-and-credentials) — the trust model behind every tool on this page * [Agent reference](/agents-lite/agent-reference) — what each autonomous module posts to Slack on its own schedule, separate from this conversation