Verbasil, company memory
आपकी टीम और उसके AI एजेंटों के लिए कंपनी मेमोरी: यह निर्णयों, सीखों और ग्राहक संकेतों को बनाए रखती है, उन्हें उनके साक्ष्य से जोड़ती है, और जो बार-बार सामने आता है उसे उजागर करती है।
दस्तावेज़
Verbasil: company memory for you, your team and your agents
Verbasil is a company memory: a service that retains an organization's decisions, learnings and customer signals, ties them to their evidence, and surfaces what keeps coming back.
Verbasil is a memory engine, not a search engine. It turns emails, meetings, notes, customer feedback and documents into decisions, learnings, signals and open loops, each tied to its evidence. It is not a wiki or a document base, not a CRM, not meeting notes, and not the memory of a third-party AI.
Hosted in Europe. Imported data is never used to train a model. French version: https://verbasil.com/fr
Pricing
- Solo: €49 per month, 1 user, personal memory + 1 shared memory
- Team: €195 per month, 5 seats included, then €39 per additional seat
- Enterprise: on request, from 50 people, SSO/SAML and SLA
When to use Verbasil
Reach for Verbasil when the job depends on what a specific company already decided, learned or heard from its customers:
- Before advising or drafting for a company, load its context: what it does, who it serves, what is settled.
- Before proposing a price, a target, a rule or a process, check it against the decisions already made, with their rationale and evidence.
- When asked "why did we decide this?", retrieve the decision with the evidence it rests on.
- When preparing a call or a message to a person, read that contact's history first.
- When a meeting, an email or a conversation produced a decision or a lesson, propose it to the memory. It waits for human approval in Sources before it counts.
- When a team wants to know what keeps coming back (objections, requests, problems), read the recurrences ranked by evidence.
Do not use Verbasil for general web knowledge, for documents a team only needs to store (it is not a drive or a wiki), or to change data without a human approving it.
For AI agents and developers
- MCP server (Streamable HTTP): https://verbasil.com/api/mcp
- MCP server card: https://verbasil.com/.well-known/mcp/server-card.json
- OAuth 2.1 protected resource metadata (RFC 9728): https://verbasil.com/.well-known/oauth-protected-resource
- OAuth authorization server metadata (RFC 8414): https://verbasil.com/.well-known/oauth-authorization-server
- Scopes:
read(context, decisions, learnings, recurrences, contacts, plans) andwrite(adds proposals; nothing is applied without human approval) - Auth: OAuth 2.1 with PKCE (S256) and dynamic client registration, or a bearer token generated in Verbasil under Integrations
- How to call it: start each conversation with
verbasil_get_project_context, search with the read tools before stating a fact, callverbasil_check_decision_conflictbefore an action that could contradict a settled decision - What a connected AI can do
- Connect Claude and connect ChatGPT
- My MCP stopped responding