Okki Go Installation, Permissions, and the RevOps Evaluation Checklist Most Teams Skip

The Short Version

Okki go is okkigo's browser-side prospecting client. Installation is roughly a four-minute process — you install a Chrome extension, sign in, complete OAuth consent, and pick which accounts (email, calendar, social) you want the agent to work against. The permissions it asks for fall into three buckets: email send/read-single-thread, calendar read/write for scheduling, and social profile access tied to whichever network you connect. It does not ask for your mailbox password, does not request broad inbox scraping, and does not require permanent CRM write access.

If you lead revenue operations and you're deciding whether to add it, the real question isn't "okki go vs manual research." It's whether the combination of data freshness + consent provenance + handoff cost beats one junior SDR building lists by hand each week. For teams sending under 400 outbound touches a month, usually not. Above that, the math typically flips — but only if you've priced the hidden side of the stack. More on that below.

Why I'm Writing This — And What I Actually Compared

I run procurement at a 90-person B2B services firm. I've managed our software budget — about $180,000 annually across sales, marketing, and revops tooling — for the past six years, and I've negotiated with 30+ SaaS vendors in that window. Every invoice lives in our cost tracking system, and I keep a running TCO sheet that (honestly) has outlasted three CRM migrations.

In Q3 2024, we evaluated four outbound platforms in parallel: one legacy data provider, one LinkedIn-automation tool, one AI SDR product, and okki go. I won't name the three I dropped — that's not the point. The point is every single one of them quoted a per-seat price that looked clean, and every single one of them had a different definition of "contact." That discrepancy alone cost us about six hours of spreadsheet reconciliation across two weeks.

Everything I'd read about agent-native prospecting said the differentiator was data coverage. In practice, for a mid-market services team, the differentiator was how much of the workflow stayed inside the tool versus how much leaked into Slack threads and stale Google Sheets. Coverage was table stakes. Handoff friction was the actual P&L line.

Okki Go Installation — What Actually Happens

The install sequence, in order:

1. Install the extension from the Chrome Web Store (or Edge equivalent). About 12 MB. It's a browser extension, not a desktop app — nothing to uninstall from your fleet via MDM.
2. Sign in with your work account and confirm your workspace. This links the extension to your okkigo org and, if your admin has configured SSO, will respect existing session policies.
3. OAuth consent screen. This is the part procurement should actually read, not click past. We'll get to specifics in the next section.
4. Connect target accounts — email, calendar, and (optionally) your social seat. Each is a separate consent prompt. You can connect email without social, and the agent will degrade gracefully.
5. Set sequence defaults. Sending window, daily cap, reply-detection rules. Defaults are conservative — 40 sends/day per seat — which is where most teams should start.

Total time: four minutes if you have your OAuth credentials ready, closer to fifteen if your IT team requires a documented consent review first (which they should).

One thing nobody tells you: uninstall isn't clean by default. Disconnecting the extension does not automatically revoke the OAuth grants. You have to go into Google Workspace (or Microsoft 365) admin and revoke the app explicitly. Budget five minutes for that when you're testing, or you'll be explaining stale app permissions to your security team three months later.

What Permissions Does Okki Go Require?

Here's the honest breakdown. Two of these are standard. One is where you should slow down and ask questions.

Email (Gmail / Microsoft 365):

• gmail.send or Mail.Send — required to dispatch sequences.
• gmail.readonly or Mail.Read — scoped to reply detection. The agent needs to know when a prospect has responded so it can pause the sequence.
• It does not request full mailbox management, and it does not request the ability to delete messages.

Calendar:

• calendar.events — read/write for proposing meeting slots and detecting booking conflicts.
• Read-only would be the safer grant, but write access is what enables one-click booking. Trade-off worth noting in your security review.

Social (LinkedIn seat, optional):

• Session-scoped access. This is the one that should trigger a conversation with legal if you operate in the EU or UK.
• What it can do: read public profile fields, send connection requests on your behalf (rate-limited by okkigo, not by LinkedIn's ToS).
• What it can't do: post, share, or modify your profile.

I assumed "OAuth" meant uniformly scoped access across those three buckets. Didn't verify. Turned out the social grant is materially broader in practice than the email grant, even though both are labeled "OAuth." Learned to read the actual scope strings after a vendor in 2023 handed us a consent screen that quietly bundled drive access into a "document sharing" permission. Different product, same lesson.

Under GDPR Article 6 and CCPA, consent provenance matters as much as consent itself. Ask the vendor where the underlying contact data was sourced and whether the lawful basis travels with the record. If they can't answer in under 90 seconds, that's your answer.

Lead Generation and Intent Data — What You're Actually Buying

Two features, one budget line, very different maturity curves.

Lead generation capabilities — the practical version is: waterfall enrichment (multiple data vendors queried in sequence, first valid match wins) plus email verification, plus a LinkedIn-side signal layer. What that means for your team is that a prospect record arrives with a confidence score, not just a green checkmark. The waterfall approach means fewer empty cells, but each cell carries a slightly different trust profile depending on which vendor filled it.

Intent data features — this is where procurement should slow down. Intent data is a category, not a feature. There are three flavors in the market (first-party, third-party aggregated, and second-party partnership data), and the only one with clean provenance is the first. Okki go leans on a mix; the mix matters more than the volume of signals.

Here's the counterintuitive bit: more intent signals is not better. We tested two datasets in Q4 2024. The one with 40% fewer signals actually converted 22% better on reply-to-meeting, because the noise floor was lower. Turns out signal density was inversely correlated with signal precision for our ICP. Nobody's marketing page mentions that.

What RevOps Teams Should Evaluate in Email Outreach

Forget feature matrices. Six items, in priority order:

1. Reply-to-meeting conversion, not reply rate. Reply rate is vanity. If your tool reports it prominently, they're selling to the wrong buyer.
2. Deliverability guardrails. Daily send caps, warm-up logic, bounce thresholds. Ask what happens when their own system hits a deliverability problem — do they pause your sequences automatically?
3. Data handoff cost. How many clicks to export a sequence into HubSpot or Salesforce? Ours was fourteen before we pushed back. Now it's three. That gap cost us roughly $4,000 in SDR time over a year.
4. Verification transparency. Not the accuracy claim — the method. SMTP handshake vs. catch-all guesses vs. managed verification. Different methods, different failure modes.
5. Intent data source breakdown. First-party only? Aggregated only? A labeled mix? Unlabeled mix means walk away.
6. Exit cost. Can you export every contact, every sequence, and every reply thread in a machine-readable format? If the answer involves "contact support," add 40 hours to your switching cost model.

The TCO math that surprised me: across four vendors we priced in 2024, list price variations were within 12%. Total cost over 18 months (including onboarding, integration, admin time, and one failed re-implementation) varied by 47%. The cheapest quote was the second-most-expensive in TCO. The most expensive quote was actually mid-pack. "Cheap" almost never survives contact with your finance system.

When Okki Go Isn't the Right Fit

Three honest boundaries.

If your outbound volume is under 300 touches/month, the per-seat economics don't clear. A skilled SDR with a decent research checklist will out-perform any AI SDR at that scale — not because the tool is bad, but because setup, QA, and sequence tuning eat the savings.

If your ICP requires deep customization per account (regulated industries, enterprise ABM, anything with a 90-day sales cycle), the agent's strength — speed — works against you. You want depth, not throughput.

If your security team hasn't signed off on browser-extension OAuth patterns, don't force it. Run the consent review first. It's a 90-minute conversation that prevents a 9-month remediation later.

Pricing below is directional and as of the current quarter — verify against current vendor quotes before you build a budget line. Never anchor a purchase decision on a number you read in an article, including this one.

I'd rather spend ten minutes explaining OAuth scopes to you than have us both discover the mismatch in month three. An informed buyer asks better questions and switches vendors less often. That's not a sales pitch — it's the cheapest form of risk management I've found in six years of running this budget.