Empirical memory for AI agents — patterns learned from real executions, not hand-written rules. Workflow checkpointing, failure knowledge base, multi-agent coordination. Zero API keys required.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"axiom-perception-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# axiom-perception-mcp **Persistent memory and pattern learning for AI agents.** Claude forgets how to use your computer between sessions. axiom-perception-mcp fixes that — it gives Claude a long-term memory for multi-step workflows so it never re-learns the same thing twice. ``` claude: recall_pattern("post tweet") → "Found pattern v3 (97% success rate, 42 executions): 1. Navigate to x.com 2. Click 'Post' in the left sidebar 3. ..." ``` Zero API keys. No setup. Just install and Claude starts remembering. --- ## Install ```bash # Claude Desktop / Claude Code uvx axiom-perception-mcp ``` Add to your MCP client config: ```json { "mcpServers": { "perception": { "command": "uvx", "args": ["axiom-perception-mcp"] } } } ``` --- ## The problem it solves When Claude tries to do something in your browser — post a tweet, create a PR, fill a form — it often spends 5-10 minutes figuring it out through trial and error. Every session, from scratch. axiom-perception-mcp stores what worked. Next time, Claude skips straight to the answer. **Works alongside any automation tool**: Playwright MCP, Computer Use, browser tools. This MCP handles *what to do*, your automation tool handles *how to do it*. --- ## 8 tools | Tool | What it does | |------|-------------| | `recall_pattern(task, app?)` | Get the best known workflow **before** starting a task | | `save_pattern(task, steps, app, category)` | Save a workflow that worked | | `update_pattern(id, steps, reason?)` | Improve a pattern with a better approach | | `record_outcome(id, success, time_ms?)` | Track executions to build success rate | | `list_patterns(app?, category?)` | Browse all known workflows | | `search_patterns(query)` | Search across all patterns | | `export_pattern(id)` | Export as JSON to share with the community | | `fetch_community_patterns(app?)` | Import proven patterns from the shared database | --- ## Community patterns — no cold start On first use, run `fetc…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.