Indicative comparison on a few axes. Products evolve quickly — check each vendor’s site before you decide.
| Capability / angle | Akasha | OpenClaw | Claude Code | Cursor |
|---|---|---|---|---|
| Primary focus | Personal assistant daemon (24/7), TUI + web + optional channels | Do-things assistant; CLI + many messengers (Slack, Telegram, WhatsApp, …) | Software development inside editor/terminal (Anthropic ecosystem) | AI-assisted editing and agents inside the Cursor IDE workflow |
| Default LLM | Embedded small model (no account); Ollama / APIs optional | Bring your own API keys or local (e.g. Ollama) | Claude models (subscription / API) | User- or vendor-configured models (see Cursor pricing/docs) |
| Runtime stack | Rust binaries + optional Tauri UI | Node-based CLI (typical install) | IDE / terminal integration | Desktop IDE (Electron) + optional cloud-backed features |
| Tool / file access | tools_policy.yaml — deny-by-default paths & commands |
Configurable (full access or sandbox per docs) | Repo-scoped coding tasks | Workspace rules, MCP, and project-scoped tools (per vendor docs) |
| Sensitive action governance | Permission review queue (pending/approved/denied/expired) + persistent decisions + task-level operator reports (GET /api/tasks/:id/report) |
Workflow-dependent | Vendor workflow and account controls | IDE/vendor controls and team policies (varies by plan) |
| Skills / extensions | SKILL.md skills + WASM plugins; install via chat or files; catalog on site (v0.7+); Cookbook recipes + Compare + mission/graph tools (v0.8+) | Skills & plugins ecosystem | MCP, slash commands, project context | Rules, MCP servers, and editor-integrated agent features |
| Observability | Timeline & metrics HTTP API, plugin list/reload/reputation (v0.7+), autonomous mission & workspace-graph APIs (v0.8+), Router tab in TUI | Dashboard / logs (varies by version) | IDE-integrated output | In-IDE chat, logs, and vendor dashboards (varies) |
| Workspace operator UI (v0.9+) | Cookbook — model pulls + 12 recipes. Compare — blind A/B. Deep research + Notes panels. Multi-page doc in Doc tab. | Workflow-dependent | IDE-scoped coding workflows | Editor-integrated agents (varies by release) |
| Long-running goals & project context (v0.8+) | Autonomous mission — YAML + heartbeat + Mission tab. Project graphs — multiple indexed roots, automatic context + workspace_graph_search, HTML reports per workspace. |
Workflow-dependent | Repository-scoped coding context | Mostly session- and workspace-scoped (check current product docs) |
| Parallel coding workflow (studio) | Code Studio swarm MVP shipped: one coordinator + bounded workers + cockpit graph states | Depends on stack/tools chosen by user | Primarily single-agent IDE coding loop | Depends on Cursor release (multi-file edits, background agents — verify on vendor site) |
| Source code | Public on GitHub (Apache-2.0); official binaries also on this site | Public repositories (open source) | Closed source (vendor) | Closed source client (vendor) |
| Account required | No (for embedded path) | No for OSS; APIs billed separately | Yes (Anthropic) | Yes for typical cloud-backed usage (see vendor) |
| Embedded GPU inference (Windows) | llama-cpp + CUDA zip; GGUF download via wizard or CLI; full CUDA bundle v0.10+ | Via Ollama / local stack (varies) | Cloud-first | Cloud-first |
| License (typical) | Apache-2.0 — github.com/azerothl/Akasha | Open source (e.g. MIT per project) | Commercial product + ToS | Commercial product + ToS |
Akasha downloads and docs live on this site and GitHub (Akasha_app). This table is not exhaustive and does not imply endorsement.
Hermes Agent emphasizes toolsets, MCP, webhooks, and operator-oriented workflows.
Akasha tracks multi-product parity in spec/dev/roadmap/reference-products-parity-matrix.md (versioned; legacy filename hermes-akasha-parity-matrix.md redirects there).
On-site guides: Docs → Operator parity.
| Theme | Akasha (indicative) | Hermes (indicative) |
|---|---|---|
| MCP | JSON validation + stdio probe (akasha mcp); OAuth hardening documented for roadmap. |
Strong MCP-first operator story; see Hermes docs and releases. |
| Webhooks / automation | Signed inbound routes + optional static direct response; scheduler + task_run APIs. |
Event-driven integrations and “direct delivery” patterns in product notes. |
| Toolsets | tool_profiles in tools_policy.yaml, CLI akasha toolset, GET /api/tools/effective. |
Named toolsets and presets per platform/session. |
| Terminal / long jobs | Shell tools + background sessions + process watch API; PTY interactive roadmap. | Multiple terminal backends and long-running job UX (see Hermes features overview). |
| Ecosystem repos | Static site (here), skills/plugins catalogs, Code Studio, Rbitnet for local inference — each aligned with the same matrix. | Hermes distribution model differs; compare vendor docs before choosing. |
Mercury Agent is an open-source, CLI-first assistant with optional Telegram access, permission-hardened tools, token budgets, and a local “second brain” memory story — check its README for the exact feature set as it evolves.
| Theme | Akasha (indicative) | Mercury Agent (indicative) |
|---|---|---|
| Primary surface | TUI, Tauri desktop, HTTP API, Code Studio web cockpit | CLI and optional chat channels (per upstream docs) |
| Default intelligence path | Embedded small model without cloud account; external providers optional | Bring-your-own models / APIs (see project documentation) |
| Permissions & cost | tools_policy.yaml, vault, human-in-the-loop; router metrics |
Explicit permission layers and token budget controls (per upstream) |
| Memory & recall | Short/long-term memory, recall metrics API, mission state (v0.8+) | Local SQLite-oriented “second brain” narrative (verify in README) |
| Distribution | Apache-2.0 engine on GitHub; official binaries + static ecosystem repos | Open source repository; self-hosted operation |